THE ROBOTIC GAIT REHABILITATION TRAINER

Size: px
Start display at page:

Download "THE ROBOTIC GAIT REHABILITATION TRAINER"

Transcription

1 THE ROBOTIC GAIT REHABILITATION TRAINER A Dissertation Presented by Maciej Dariusz Pietrusinski to The Department of Mechanical and Industrial Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the field of Mechanical Engineering Northeastern University Boston, Massachusetts April, 2012 i

2 Abstract Current methods of robotic neurorehabilitation of gait often do not address secondary gait deviations, focusing instead on only the primary gait deviations. Therefore, a robotic system was developed, which guides the pelvis in the frontal plane (pelvic obliquity), in order to address hip-hiking - the most common secondary gait deviation. A prototype of the device was built with a single actuator and impedance control system to generate force field and transfer it to the patient s pelvis via a lower body exoskeleton. The RGR Trainer s ability to alter gait pattern via force fields applied to pelvic obliquity was tested on several healthy subjects. It was found that the RGR Trainer can coax healthy subjects to walk with an altered gait pattern, and signs of retention of this newly learned gait pattern have been observed. Thesis Supervisor: Prof. Constantinos Mavroidis Professor of Mechanical Engineering ii

3 Acknowledgement I owe my deepest gratitude to my advisor, Professor Constantinos Mavroidis, who has given me the opportunity to work under his guidance at the Biomedical Mechatronics Laboratory, and on the research project presented in this thesis. I am also very grateful to Dr. Paolo Bonato from Spaulding Rehabilitation Hospital, for his involvement in the project and the human subject testing of the RGR Trainer. I m also very grateful for the time and energy invested by Iahn Cajigas from Spaulding Rehabilitation Hospital, who has been involved in many aspects of this project from the very beginning. I would also like to thank my colleagues at the Biomedical Mechatronics Laboratory: Ozer Unluhisarcikli, Richard Ranky, Mark Sivak and Brian Weinberg, who have been all very supportive throughout the duration of my graduate studies, and who have contributed in many ways to the project. Finally, I d like to thank my parents, Jan and Grażyna for their emotional and financial support throughout my life and during the last 5 years of my graduate studies in particular. The Robotic Gait Rehabilitation Trainer project presented in this thesis has been funded by the National Science Foundation (NSF) Grant iii

4 Biographical Note The author received his BS in Mechanical Engineering from University of Massachusetts at Amherst in December of He entered the department of Mechanical and Industrial Engineering at Northeastern University in September of 2007, and he began working at the Biomedical Mechatronics Laboratory under prof. Mavroidis on the NSFfunded project Pelvic Obliquity Rehabilitation in Stroke Patients Using Robotically Generated Force-Fields in the summer of He received his MS in Mechanical Engineering in August of iv

5 Table of Contents Abstract... ii Acknowledgement... iii Biographical Note... iv List of Figures... vii List of Tables... xiii Glossary... xiv Chapter 1. Introduction Problem Description Significance Contributions Overview... 6 Chapter 2. Background Human Gait Pelvis Motion during Gait Common Gait Deviations in Pelvic Motion Gait Rehabilitation Robotic Gait Rehabilitation Conclusion Chapter 3. On the Mechanical Design of RGR Trainer Introduction RGR Trainer Working Principle RGR Trainer Mechanical System Overview Actuation System Human - Robot Interface Conclusion Chapter 4. RGR Trainer Control System Introduction Impedance Control Theory Actuator Position Feedback Force Feedback Control Hardware and Software Force Controller Tuning v

6 4.7 Linear Motion Impedance Controller Bench Tests Pelvic Obliquity Position Feedback Pelvic Obliquity Impedance Controller Human Machine Synchronization Overall Control System Architecture Actuation System Backdrivability Actuation System Bandwidth Safety Conclusion Chapter 5. Healthy Subject Testing Introduction Protocol Protocol Protocol Protocol Conclusion Chapter 6. Conclusions Summary Future Work Appendix A - Impedance Controller Bench Test Results Appendix B - Protocol 3 Results Appendix C - RGR Trainer 2DOF Bibliography vi

7 List of Figures Figure 1: The human gait cycle [7] Figure 2: The body can be viewed in the frontal, sagittal and transverse planes [8] Figure 3: Normal pelvis motion events in gait [9] Figure 4: Hip hike is a voluntary upward motion of the contralateral (affected) side of the body during leg swing [9] Figure 5: Circumduction is used by subjects to create additional foot clearance Figure 6: Manual treadmill gait retraining is labor intensive and physically demanding. Image adapted from [15] Figure 7: The Lokomat in action [18] Figure 8: LOPES nine degrees of freedom (eight are actuated) [19] (1) foreward-back translation, (2) lateral translation, (3) vertical translation, (4) hip abduction, (5) hip flexion, (6) knee flexion Figure 9: LOPES structure [19] Figure 10: The HapticWalker [20] Figure 11: PAM and POGO [6] Figure 12: With one leg in swing, a moment can be applied onto the pelvis about the weight supporting hip joint with just one actuator. Adapted from [9] Figure 13: Subject in the RGR Trainer with major components labeled Figure 14: Servo tube linear actuator from Copley Controls Inc Figure 15: Sensing (pelvis orientation and force) and actuation in the RGR Trainer. The servo-tube actuator (which contains hall-effect sensor based internal position measurement) and the linear potentiometer are fixed to the frame of the RGR Trainer, but can follow the motion of the body in the horizontal plane Figure 16: Horizontal motion system upgrade with the actuation system. Triangular subassemblies support the linear actuator assembly and the linear potentiometer assembly. A revolute joint about the vertical axis and a prismatic joint in the horizontal plane provide unconstrained motion in the horizontal plane while constraining motion in the vertical direction Figure 17: The horizontal motion system (upgrade) with the actuation system is secured to the Biodex frame with four locking clamp subassemblies. Body weight support components are not shown Figure 18: Revolute joint detail showing one tapered roller bearing mounted on a precision shaft. Two opposing bearings support axial loads in either direction, and radial loads. Set screws are used to lock the mounting blocks to the precision shaft and keep the distance between the bearings fixed Figure 19: Complete horizontal motion system ready to be mounted on the Biodex frame Figure 20: Newport 4 pelvic brace with thigh segments attached Figure 21: LOPES DOFs, including free hip flexion/extension and free abduction/adduction Figure 22: BLEEX worn by a user Figure 23: Pelvic brace design of BLEEX exoskeleton Figure 24: Complete human-robot interface suspended from the RGR Trainer s actuation system front view Figure 25: HRI top view. Plastic shells wrap around subject s pelvis Figure 26: Right side of the human-robot interface, with all DOFs (left) and adjustments (right) shown. Plastic shell interfacing with subject s waist was removed for clarity. The DOF axes are: (1) hip flexion, (2) hip abduction, (3) hip internal/external rotation, (4) knee flexion, (5) ankle flexion. Adjustments: (a) hip joint span, (b) pelvis width, (c) thigh length, (d) shank length, (e) knee frontal plane angle vii

8 Figure 27: Hip revolute joint, with potentiometer for flexion-extension angle measurement. Precision shaft is double supported by a pair of needle pin bearings and thrust bearings Figure 28: Knee joints with adjustable frontal plane angle (two extremes shown) and rotary potentiometer for knee flexion/extension measurement. The design is optimized to resist moments in the frontal plane, resulting from force fields applied to pelvic obliquity Figure 29: Basic outline of impedance control architecture. Proportional and derivative gains (PD) produce a force command which is executed by the force loop with gain G. System s interaction force with the environment (F ext ) is measured with load cell Figure 30: Simple model of actuator s thrust rod Figure 31: Actuator shaft and force control law Figure 32: Dynamics of the actuator Figure 33: Physical implementation of equation (4.14) Figure 34: The unconditioned load cell signal contains significant noise Figure 35: Once analog filtered, the noise level in the signal is greatly reduced Figure 36: Unfiltered load cell signal sinusoidal loading at 1Hz Figure 37: Low pass analog filtered (480Hz cutoff) load cell signal sinusoidal loading at 1Hz Figure 38: Combination of analog filtering to remove high frequency aliases, and digital filtering produces a clean force signal Figure 39: Xenus servo amplifier from Copley Controls Inc Figure 40: Outline of the inner current loop contained in the Xenus servo amplifier Figure 41: Actuator shaft coupled to the body via pelvic brace, with load cell reading the interaction forces. Figure 42: In open loop mode, we see about 15-20% steady state error Figure 43: Closed loop step response with proportional gain G= Figure 44: Closed loop step response. With a proportional gain of 1.8, serious instability occurred Figure 45: Linear motion impedance controller was used in bench testing Figure 46: Position error (Des. Position Act. Position) generated the force command (F virt ). Load cell measured actual interaction force (F ext ) Figure 47: The addition of damping attenuated the oscillatory force interaction. Effects of stiction can be seen just past maxima and minima Figure 48: Higher gain value caused greater environment deflection (Act. Position). Lack of damping resulted in oscillatory response Figure 49: After the damping ratio (zeta) was introduced, the oscillatory behavior diminished Figure 50: At this high virtual stiffness setting, slight vibrations were again felt, and can be seen in the F ext signal Figure 51: Increasing the damping ratio to 0.8 amplified the vibrations Figure 52: Obliquity angle can be calculated knowing vertical position of the two attachment points. D is the length of the direct line between the two attachment points, and y is the distance between them in the vertical direction Figure 53: The PD controller acts on the obliquity error and outputs the appropriate force command. Low pass filters 1 and 2 are RC anti-alias filters Figure 54: Details of the PD gain block from Figure 53. The proportional gain K c is specified at the obliquity level, while the derivative gain B c acts on linear velocity error at the actuator level. B c is computed from K c (linear motion equivalent) and the specified damping ratio ζ using Equation Velocity feedback undergoes secondary filtering (after velocity error is computed) Figure 55: Conceptual diagram and synchronization algorithm diagram, adapted from [6] Figure 56: Foot switch construction. Clear plastic sheet taped over the top improves user comfort viii

9 Figure 57: MATLAB s unwrap function produces continuous curves of periodic time series. The range of phase angle of 5 radians to 75 radians covers approx. 11 full gait cycles Figure 58: With offset introduced to remove delay, heel strikes as predicted by gait estimation algorithm ( T rep ) nearly coincide with those produced by the discrete gait event (heel strike) Figure 59: Overall Control System Architecture Figure 60: Two consecutive gait cycles. Synchronization algorithm output predicts left heel strikes well, and gives good estimate of gait cycle location mid-stride. Gait estimation (Synchr Output) is the progression through the gait cycle from 0 to 1 (100%). Force field activation sigmoid switch (3Hz) was set to go on at 44% and off at 76%. Heel strike is marked by the rising edge of the Heel Strike Switch signal Figure 61: Layout of hardware components of the RGR Trainer s control system Figure 62: Interaction force data with the RGR Trainer s control system set to follow mode, under two force gain settings, collected at 50Hz Figure 63: Backdrivability test results (with force control gains as indicated). The healthy subject ambulated at his comfortable walking speed (CWS) of 3km/h. Approximately 50% (6dB) interaction force reduction for frequencies 0-6Hz can be seen Figure 64: Bandwidth test setup with two compression springs (k=5.66kn/m) Figure 65: Commanded force (Chirp Force Command) and the resulting interaction force Figure 66: Actuator force bandwidth test results Figure 67: Analog amplifier-enable safety circuit Figure 68: Pelvic obliquity trajectories collected from healthy and impaired subjects in the study by Cruz et al. [44]. The impaired subjects (stroke) clearly exhibit a hip-hiking pelvic motion trajectory Figure 69: Graphical representation of protocol Figure 70: Result of hip-hike inducing test. The interaction force magnitudes measured by the load cell (F int ) are graphed. Models fitted to the data (Fit) were used to estimate time constants Figure 71: Graphical representation of protocol Figure 72: Sample result from one subject tested under Protocol 2. In the top graph, ref BL is the baseline pelvic obliquity of the subject, ref HH is the hip-hiking trajectory from Cruz et.al, act BL is the mean pelvic obliquity trajectory under force field, act HH is the average hip-hiking trajectory produced by the subject under force field and act AE is the average pelvic obliquity following hip-hike training session. In the bottom graph, pelvic obliquity curves from all gait cycles in epoch 4 are shown (dashed blue) along with baseline (solid black). Here one gait cycle spans between consecutive left foot toe-offs Figure 73: Subject # 2 baseline and hip-hiking plots. Baseline and its standard deviation curves are plotted along with their inverses (with first half of gait cycle plotted first, and vice versa) in order to aid in visualizing symmetry. Right hip-hike curves are also inverted and plotted in reverse order to facilitate comparison with left hip-hike curves Figure 74: Subject # 3 baseline and hip-hiking plots Figure 75: Subject # 4 baseline and hip-hiking plots Figure 76: Subject # 7 baseline and hip-hiking plots Figure 77: Subject # 8 baseline and hip-hiking plots Figure 78: Mean pelvic left hip-hike obliquities of 8 subjects (across ca. 200 gait cycles) and mean across the means of 7 subjects (subject # 2 data excluded due to the time-series extreme mismatch with the rest due to erroneous foot switch operation) Figure 79: Graphical representation of protocol 3 trial. A session consisted of five such trials concatenated into a single run. Epoch 1 was only used in trial 1 of the run, and each trial used a different force field magnitude ix

10 Figure 80: Taking into account the offset in pelvic obliquity between baseline (ref BL ) and pelvic obliquities resulting from hip-hike training, an after-effect can be observed (act AE-Early ) which diminishes with time (act AE-Late ) Figure 81: Subject 2 exhibits significant after-effect, characterized by exaggerated pelvic drop at the beginning of the third epoch (backdrive), labeled as act AE-Early. This after-effect diminishes throughout the duration of the third epoch, and the de-adapt epoch (4 th ) seems to accomplish its task (act de-adapt ) Figure 82: This subject s baseline pelvic obliquity (ref BL ) has a gross offset. Nevertheless, the action of the RGR Trainer does make the subject produce a hip-hike during the second half of the gait cycle. Unlike what was observed in Figure 80 and 81, the subject exhibits a reduced pelvic drop, which returns to baseline over time Figure 83: Graphical representation of a single trial of protocol 4. A complete session consisted of three trials running continuously (each using a different force field), with trial 2 and 3 consisting of epochs 2 through 5. Epoch 3 type was randomized. Two sessions per training type (assistive or resistive) were run to ensure that every combination of force field and epoch 3 type was tested Figure 84: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at K c = 5N-m/deg) followed by epoch 3b (backdrive, check-). Subjects hip-hiked during the training period, and exaggerated pelvic drop in the subsequent epoch may indicate motor adaptation Figure 85: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 15N-m/deg) followed by epoch 3b (backdrive, check-). Subjects hip-hiked during the training period. The exaggerated pelvic drop in the subsequent epoch (3b) may be a sign of motor adaptation Figure 86: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 25N-m/deg) followed by epoch 3b (backdrive, check-). Subject 2 exhibited most exaggerated pelvic drop in epoch 3b Figure 87: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 5Nm/deg) and to baseline reference ( strides, epoch 3b) during assistive training Figure 88: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 15Nm/deg) and to baseline reference ( strides, epoch 3b) during assistive training. Only subject 4 exhibited a gait pattern close to hip-hiking for several gait cycles after the switch occurred Figure 89: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 25Nm/deg) and to baseline reference ( strides, epoch 3b) during assistive training Figure 90: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 5N-m/deg) followed by epoch 3b (backdrive-check) Figure 91: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 15N-m/deg) followed by epoch 3b (backdrive-check) Figure 92: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 25N-m/deg) followed by epoch 3b (backdrive-check) Figure 93: Resistive training at 5N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp) Figure 94: Resistive training at 15N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp). Subject 1 resisted the force field the most, peaking at almost 100N (mean), and during the Error Clamp epoch he again fought against the system the most (-40N mean). This particular subject reported verbally during the training session having difficulty recalling his baseline pelvic obliquity pattern x

11 Figure 95: Resistive training at 25N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp) Figure 96: Position error (Des. Position Act. Position) generated the force command (F virt ). Load cell measured actual interaction force (F ext ) Figure 97: The addition of damping attenuated the oscillatory force interaction. Effects of stiction can be seen just past maxima and minima Figure 98: Higher gain value caused greater environment deflection (Act. Position). Lack of damping again resulted in oscillatory response Figure 99: Once again, after the damping ratio (zeta) was introduced, the oscillatory behavior diminished Figure 100: At this high virtual stiffness setting, slight vibrations were again felt, and can be seen in the F ext signal Figure 101: Again, with the damping ratio increased, the vibrations diminish Figure 102: As the damping ratio was increased to 0.6, undesirable behavior appeared. The virtual damper component of the command signal began displaying vibratory behavior. The resulting forces were felt by the subject, but are not present in the measured force signal F ext due to low-pass filtering Figure 103: With the reference trajectory of 3Hz and no damping, the measured force signal (F ext ) tended to lag behind the commanded force (F virt ) Figure 104: Increase in damping ratio smoothed out both force curves Figure 105: With the damping ratio set to 0.5, the system still behaved well Figure 106: Once the damping ratio was increased to 0.6, the performance deteriorated due to appearance of high frequency vibrations, which can be seen in the F virt signal, and could be felt by the subject Figure 107: With the reference trajectory frequency increased to 6Hz, actuator thrust rod s inertia caused significant distortions to the position and force profiles. The system still behaved in a stable manner Figure 108: Introduction of damping had the effect of correcting the profile of the external measured force F ext, by properly modulating the virtual force F virt Figure 109: Inertial effects cause the measured force profile F ext to lag significantly behind position error Figure 110: Increasing the damping ratio seemed to make the controller efforts (F virt ) more abrupt Figure 111: As seen before, damping ratio of 0.6 amplified the derivative action of the PD impedance controller. This caused high frequency vibrations Figure 112: Subject 1 2 nd trial Figure 113: Subject 1 3rd trial Figure 114: Subject 1 4th trial Figure 115: Subject 1 1st trial Figure 116: Subject 1 5 th trial Figure 117: Subject 2 5 th trial Figure 118: Subject 2 2 nd trial Figure 119: Subject 2 1 st trial Figure 120: Subject 2 3 rd trial Figure 121: Subject 2 4 th trial Figure 122: Subject 3 1 st trial Figure 123: Subject 3 2 nd trial Figure 124: Subject 3 3 rd trial Figure 125: Subject 3 4 th trial Figure 126: Subject 3 5 th trial Figure 127: Control link optimized for mass xi

12 Figure 128: Concept 2 with torsion bar applying moments in pelvic obliquity and two push-rods in pelvic rotation, here pictured with two Copley linear actuators Figure 129: Constant-radius arc guided by bearings used to place the center of rotation within the subject s body Figure 130: The gimbals at the hip joints allow for hip abduction/adduction and flexion/extension. Flexible transmission is required to apply moments at pelvic obliquity Figure 131: RGR Trainer 2DOF actuation concept Figure 132: Actuator mount detail design Figure 133: Closed linkage mechanism Figure 134: Top view of the mechanism Figure 135: Definition of pelvic obliquity Φ and pelvic rotation θ angles. Pelvic brace as viewed from behind Figure 136: Forces f L and f R necessary to produce desired net forces and moments Figure 137: Force f resolved into the component magnitudes and unit vectors Figure 138: RGR Trainer 2DOF frame structure only. BWS beam is reinforced with steel cables Figure 139: Side view of RGR Trainer 2DOF over treadmill. Manipulators are shown attached to pelvic interface worn by 99 th percentile female subject Figure 140: Manipulator attachment to frame Figure 141: Handle bar tilt adjustment using quick release clamps and spring loaded plunger Figure 142: Handle bar height adjustment xii

13 List of Tables Table 1: Summary of hip-hike data. The means exclude subject # 2 data. Only one subject (#6) was left-foot-dominant xiii

14 Glossary Afferent - Brain plasticity - Central Pattern Generators - Contralateral - Equinus - Hemiparesis - Ipsilateral - Motor adaptation - Motor cortex - Motor learning - Neurorehabilitation - Orthosis - anatomical term: towards the center of the body the capacity of the nervous system to change its structure and network, neurogenesis, its cognition and function over a lifetime neural networks that produce rhythmic patterned outputs without sensory feedback side of the body opposite to the side of brain lesion; generally this side of the body is the one affected by stroke condition characterized by tiptoe walking on one or both feet. It is usually associated with clubfoot weakness on one side of the body same side of the body as the side of brain lesion; generally this side of the body is not affected by the stroke modification of a movement from trial to trial based on error feedback region of the cerebral cortex involved in the planning, control, and execution of voluntary motor functions formation of new motor pattern that occurs via long-term practice (i.e. days, weeks, years) complex medical process which aims to aid recovery from nervous system injury, and to minimize and/or compensate for any functional alterations resulting from it orthopedic appliance or apparatus used to support, align, prevent, or correct deformities or to improve function of movable parts of the body xiv

15 Introduction 1.1 Problem Description Each year 800,000 people suffer a stroke in the United States alone [1]. Stroke is a leading cause of disability. Stroke survivors experience weakness and difficulties moving one side of the body (i.e. they are affected by hemiparesis), with a negative effect on the performance of motor activities such as walking. Walking allows individuals to perform activities of daily living [2, 3]. The ability to walk is strongly correlated with quality of life [4]. Hemiparesis and abnormal synergy patterns are characteristic of gait disorders following stroke. Abnormal synergy patterns include equinus synergy, paretic synergy and reflex coactivation [5]. Comfortable walking speed is reduced in stroke survivors. Asymmetries mark post-stroke ambulation. Asymmetry of stance time during gait, a common feature following stroke, often limits walking efficiency, results in instability, and causes an aesthetically sub-optimal gait pattern. Therefore, the restoration of a normal gait pattern is an important goal of post-stroke rehabilitation. 1.2 Significance Many rehabilitation approaches have been used to promote functional recovery in stroke survivors. Unfortunately, the rehabilitation process is labor intensive, since it often relies on a one-to-one administration of therapy, i.e. clinicians work with a single patient at a time. Robotic systems for gait retraining have been recently developed to facilitate the administration of intensive therapy. Most of the existing systems focus on the correction 1

16 of primary gait deviations, such as knee hyperextension during stance and stiff-legged gait (defined as limited knee flexion during swing). It is often assumed that secondary deviations would be no longer observed once primary deviations are corrected. Secondary gait deviations are gait abnormalities that result from compensatory movements associated with a primary gait abnormality. Secondary deviations often involve the control of the pelvis. For instance, stiff-legged gait is often associated with hip hiking and/or circumduction of the limb. Hip hiking is an exaggerated elevation of the pelvis on the affected side of the body to allow toe clearance during swing. Circumduction of the limb is marked by an exaggerated hip abduction in combination with an exaggerated rotation of the pelvis. Abnormal control of pelvic obliquity and pelvic rotation are common secondary gait deviations observed in stroke survivors. They are often present even after rehabilitation interventions addressing the primary gait deviation that they are thought to be related to (e.g. limited knee flexion during swing, namely stiff-legged gait). Existing systems for robotic-assisted gait training typically neglect gait deviations associated with an abnormal control of the pelvis. The Pelvic Assist Manipulator (PAM) is one of the few robotic devices that attempts to address such gait deviations [6]. While, the PAM is designed to control five degrees of freedom of the pelvis during gait, the method of force transfer to the pelvis to control pelvic obliquity seems to be limited in effectiveness, and authors have reported on experiments with force fields around pelvic obliquity of 3.5N-m/deg, which in light of findings presented in this thesis is rather low. The development of a device simpler than PAM and specifically designed to control pelvic obliquity and address hip-hiking in patients post stroke has been proposed by Dr. 2

17 Paolo Bonato, the director of the Motion Analysis Lab at Spaulding Rehabilitation Hospital in Boston, and Assistant Professor at Harvard Medical School. The result is a robotic device of low mechanical complexity, presented in this thesis, which allows all the natural motions of the pelvis, while being able to selectively and compliantly guide the pelvis in the frontal plane (pelvic obliquity) in order to target hip-hiking in patients post stroke. This device uses impedance control and human-machine synchronization to generate corrective forces as a response to deviations from pre-determined pelvic obliquity trajectories. The corrective force fields are applied onto the subject via a lower body exoskeleton, which can very effectively transfer forces to the pelvis, while its 10 DOFs allow for unhindered ambulation on the treadmill. 1.3 Contributions The main contributions of this thesis are: - A novel design and actuation method of a robotic device, which applies forces to the pelvic area in order to affect the pelvic obliquity angle during ambulation. - The novel design of a lower body exoskeleton, which can effectively and reliably transfer moments to the pelvis in the frontal plane. - A control method, which facilitates application of determinate moment onto the pelvis in frontal plane with a single actuator. - Demonstration of feasibility of inducing the learning of new gait patterns via application of force fields to pelvic obliquity. Some of the work presented in this thesis has been the subject of the following publications: 3

18 - Pietrusinski M., Cajigas I., Bonato P. and Mavroidis C., "Healthy Subject Testing with the Robotic Gait Rehabilitation(RGR) Trainer, Proceedings of the CISM- IFToMM Symposium on Robot Design, Dynamics, and Control, June 12 15, Pietrusinski M., Cajigas I., Bonato P. and Mavroidis C., "Robotic Gait Rehabilitation Trainer Pelvic Obliquity Trajectory Recording with Robotic Gait Rehabilitation (RGR) Trainer and Lower Body Exoskeleton, Proceedings of the Dynamic Walking Conference (DWC), May 21 May 24, Pietrusinski, Maciej; Unluhisarcikli, Ozer; Mavroidis, Constantinos; Cajigas, Iahn; Bonato, Paolo;, "Design of human Machine interface and altering of pelvic obliquity with RGR Trainer," Rehabilitation Robotics (ICORR), 2011 IEEE International Conference on, vol., no., pp.1-6, June July Pietrusinski, M.; Cajigas, I.; Goldsmith, M.; Bonato, P.; Mavroidis, C.;, "Robotically generated force fields for stroke patient pelvic obliquity gait rehabilitation," Robotics and Automation (ICRA), 2010 IEEE International Conference on, vol., no., pp , 3-7 May Pietrusinski, M.; Cajigas, I.; Mizikacioglu, Y.; Goldsmith, M.; Bonato, P.; Mavroidis, C.;, "Gait Rehabilitation therapy using robot generated force fields applied at the pelvis," Haptics Symposium, 2010 IEEE, vol., no., pp , March Submitted: - Pietrusinski M., Severini G., Cajigas I., Bonato P. and Mavroidis C., "Design of a gait training device for control of pelvic obliquity," submitted for possible 4

19 presentation in Engineering in Medicine & Biology (EMBC), 2012 IEEE International Conference on, August 28 September 1, Pietrusinski M., Cajigas I., Severini G., Bonato P. and Mavroidis C., "Robotic Gait Rehabilitation Trainer," submitted for possible publication in the IEEE / ASME Transactions on Mechatronics, March, In preparation: - IEEE Transactions on Neural Systems and Rehabilitation Engineering, June, 2012 As a result of this work, two provisional patent applications have been filed: - Pietrusinski M., Mavroidis C., "Mobile Wearable Orthopedic Lower Body Exoskeleton for Control of Pelvic Obliquity during Gait," Invention Disclosure submitted on November 30, 2011 (INV-1234). Initial provisional patent application filed on December 5, Pietrusinski M., Mavroidis C., Bonato P., Unluhisarcikli O., Cajigas I., Weinberg B., "Orthopedic Lower Body Exoskeleton for Torque Transfer to Control Rotation of Pelvis during Gait," Invention disclosure submitted on May 31, 2011 (INV-1148). Initial provisional patent application filed on June 24, The publication presented at the International Conference on Robotics and Automation (ICRA 2010) was also nominated for the best medical robotics paper award. 5

20 1.4 Overview This thesis is an attempt to address several questions. In Chapter 2 the current state of the art compliantly controlled robotic devices for gait rehabilitation are presented. Chapter 3 describes the design of a robotic device of low mechanical complexity such that it can reliably and effectively apply force fields to pelvic motion in the frontal plane (pelvic obliquity) when walking on a treadmill. Chapter 4 presents how to control The RGR Trainer such that it can generate the prescribed corrective force fields, and finally in Chapter 5 several experimental protocols are tested on healthy subjects, providing insight on the many challenges and the right approaches for gait retraining by application of force fields at the pelvis. 6

21 Background 1.5 Human Gait Human gait is comprised of strides, which are the intervals between two consecutive heel strikes (Figure 1). Gait markers, (e.g. toe off), are used to identify the phases of gait (e.g. swing phase and stance phase). Figure 1: The human gait cycle [7]. The stance phase lasts approximately 60% of the gait cycle, while the swing phase takes up the remaining 40%. Both limbs are in contact with the ground for about 10% of the cycle, which is referred to as double limb support. 7

22 1.6 Pelvis Motion during Gait During normal gait, the pelvis rotates in three planes: frontal, sagittal and transverse. Figure 2: The body can be viewed in the frontal, sagittal and transverse planes [8]. Rotation of the pelvis in the frontal plane is obliquity, rotation in the sagittal plane is pelvic tilt, and rotation in the transverse plane is called pelvic rotation. During single limb support, these rotations happen about the supporting limb s hip joint [9]. Pelvic drop, anterior tilt and rotation are normal events which occur during normal gait in obliquity, pelvic tilt and pelvic rotation respectively. 8

23 Figure 3: Normal pelvis motion events in gait [9]. 1.7 Common Gait Deviations in Pelvic Motion The most common primary gait deviation in patients post stroke is stiff legged gait. This gait deviation oftentimes results in the subject employing secondary gait deviations which involve motor control of the pelvis. Stiff legged gait is associated with hip-hiking (Figure 4) or circumduction (Figure 5). Hip-hiking is an exaggerated elevation of the pelvis on the contralateral side (i.e. hemiparetic side) to allow toe clearance during swing, while circumduction is an exaggerated rotation of the pelvis in combination with an exaggerated hip abduction. Abnormal control of pelvic obliquity and rotation of the pelvis are the most common secondary gait deviations observed in post-stroke patients. A subject will employ these secondary gait deviations in order to assist in foot clearance when either hip flexion or knee flexion are inadequate [9]. 9

24 Figure 4: Hip hike is a voluntary upward motion of the contralateral (affected) side of the body during leg swing [9]. Figure 5: Circumduction is used by subjects to create additional foot clearance. 10

25 1.8 Gait Rehabilitation Animal research studies have shown that goal oriented, repetitive training is the primary means of augmenting post-stroke motor relearning [10]. Human clinical trial studies that utilize goal-oriented, repetitive, active training such as constrained induced movement therapy [11], partial weight-supported ambulation [12] and robotic therapy [13] have demonstrated encouraging results. Based on the growing body of scientific evidence pointing to the effectiveness of goal-oriented motor retraining, clinicians have recently privileged a goal-oriented approach also in gait retraining, and have utilized treadmills to implement clinical protocols. Studies examining treadmill gait retraining have shown its effectiveness in improving walking velocity and other key characteristics of ambulation [12, 14] and a positive effect on mobility. Treadmill walking is used as a substitute to level ground walking since not only it appears to be an effective clinical tool, but also it offers some practical advantages over level ground gait retraining. For instance, treadmill gait retraining uses less space and is relatively simple to apply this technique in less functional patients with the use of weight support. Studies have shown that walking on a treadmill does not significantly change the gait pattern compared to level ground waking [14] and that improvements achieved during treadmill gait retraining transfer to level ground walking. In cases when patients are unable to properly ambulate, use of a treadmill for gait retraining makes it easier for physical therapists to administer motion to lower extremities manually (Figure 6). 11

26 Figure 6: Manual treadmill gait retraining is labor intensive and physically demanding. Image adapted from [15]. Unfortunately, there are two major drawbacks to manual therapy: it s difficult for the two therapists to coordinate their work properly, and it is labor intensive, therefore making it difficult to implement in the US healthcare system. 12

27 1.9 Robotic Gait Rehabilitation Due to the difficulties associated with manual gait retraining, robotic gait retraining systems have been developed to facilitate administration of intensive gait retraining therapy. From the point of view of training strategy and robot control, there are two types of robotic devices for rehabilitation: those which drive the body components in position mode regardless of patient efforts, and those which apply force-fields to the body, therefore modulating the forces applied onto the body depending on patient s efforts. The latter method, employing force-fields, has been shown to be the preferred method for retraining post-stroke subjects to regain their motor functions [16]. Therefore, only those robotic devices, which apply force-fields (with force measurement) to the lower body for the purpose of gait retraining, are presented here. The leader in the field of robotic neurorehabilitation is a Swiss company, Hocoma AG, which manufactures the Lokomat, a robotic device for gait retraining. The system consists of the robotic gait orthosis (Lokomat), a body weight support system (Lokobasis) and a treadmill (Figure 7). This device controls the patient s leg movements in the sagittal plane, by actuating hip and knee joints. The force fields are realized by use of impedance control. The device also features a passive foot lifter, which helps with ankle dorsiflexion in the swing phase. The subject s pelvis is fixed in the horizontal plane, but slight rotations of the pelvis are possible due to cushions and straps used to hold the body [17]. The hip and knee flexion/extension DOFs, and the vertical location of the body constitute a total of 5 actuated DOFs. This system is commercially available. 13

28 Figure 7: The Lokomat in action [18]. Another robotic device, which is designed to apply force fields in gait retraining, is the LOPES, from University of Twente in The Netherlands. This device is similar to the Lokomat, but in addition to controlling the hip and knee joints in the sagittal plane, the LOPES features additional degrees of freedom to allow pelvis translations in the horizontal plane, as well as hip joint abduction/adduction (Figure 8). The device is not available commercially. 14

29 Figure 8: LOPES nine degrees of freedom (eight are actuated) [19] (1) foreward-back translation, (2) lateral translation, (3) vertical translation, (4) hip abduction, (5) hip flexion, (6) knee flexion. Figure 9: LOPES structure [19]. 15

30 Another device, which employs force feedback in application of motion trajectories to the lower body, is the HapticWalker from the Fraunhofer Institute for Production Systems and Design Technology in Berlin, Germany. Figure 10: The HapticWalker [20]. This device is comprised of two 3 degree-of-freedom modules, which use 6 DOF force/torque sensors in its foot plates. Up to six DOF s per foot are available. The unique design of the HapticWalker allows for simulation of a wide number of trajectories like stair climbing, but it uses position control. The latest published article about this device [21] mentions force-field type control algorithms being under development. The only robotic device for gait rehabilitation, which allows all the natural motions of the pelvis while at the same time being able apply corrective moments to it is the Pelvic 16

31 Assist Manipulator (PAM) [6], shown in Figure 11 being used together with the Pneumatically Operated Gait Orthosis (POGO). PAM can apply forces to all pelvis translations and moments to pelvic obliquity and pelvic rotation, for a total of 5 actuated DOFs. Pelvic tilt is unactuated. Figure 11: PAM and POGO [6]. PAM and POGO are compliantly actuated with compressed air, making the device highly backdrivable. PAM s compliance presented new challenges not seen before in overtreadmill gait retraining robotic devices, such as the need for the machine to synchronize to the subject s stepping timing and cadence. Aoyagi et al. solved this problem by developing a synchronization algorithm, which has been implemented in many other devices ever since, including the RGR Trainer, which is presented in this thesis. Nevertheless, one major shortcoming of the PAM and POGO system is the physical 17

32 interface between the robot and the human, as the authors point out. Moments are applied to the pelvis via a semi-rigid belt, and the POGO which is not directly linked mechanically to the pelvic belt exerts flexion/extension moments at the hips and knees. The authors reported on conducting tests with spinal cord injury patients, where the force fields which PAM was configured to generate around pelvic obliquity and pelvic rotation were 200N-m/rad (3.5N-m/deg), which is quite low Conclusion The Lokomat, the LOPES and the HapticWalker systems described above have the capability to correct primary gait deviations, such as knee hyperextension during stance and stiff legged gait defined as limited knee flexion during swing, but the secondary gait deviations in the pelvis are not targeted. PAM is the only device for gait retraining which allows all natural motions of the pelvis and can apply moments to the pelvis, but its performance seems to be limited by the design of the physical interface between the robot and the subject. The development of the RGR Trainer was proposed by Dr. Paolo Bonato, to facilitate robotic gait retraining using force-fields applied to the secondary gait deviations in the pelvic motion. The RGR Trainer allows all of the natural motions of the pelvis, and features a lower body exoskeleton which employs the waist, thighs, shanks and feet to transfer moments to the pelvis. The mechanical design of the RGR Trainer coupled with the lower body exoskeleton, highly backdrivable linear actuator, impedance control and PAM s synchronization algorithm produced a gait retraining device which can effectively and reliably apply corrective moments to pelvic obliquity. These features make the RGR 18

33 Trainer arguably the best robotic system for studying motor control of pelvic obliquity in healthy subjects, which may lead to developing better gait retraining therapies for patients post-stroke. 19

34 On the Mechanical Design of RGR Trainer 1.11 Introduction Neurorehabilitation, whether in upper or lower limbs, puts forth certain desirable qualities, which robotic devices should possess, such as high backdrivability and force controllability [6]. Some devices have been developed with these qualities in mind [22-24]. On the other hand the Lokomat was first designed as a position-controlled device, and only later was it outfitted with impedance control, in order to improve its performance [25]. In light of this, the actuation system and the human-robot interface of the RGR Trainer were designed to be simple, with low moving mass and low friction, easing the task of the control system in generating appropriate performance of the overall system RGR Trainer Working Principle The RGR Trainer is a stationary device, which is placed over a treadmill, and which generates force fields around the user s pelvis, while they ambulate on the treadmill, in order to administer gait retraining therapy. The particular secondary gait deviation, which the RGR Trainer targets in patients post-stroke, is hip-hiking. Hip-hiking occurs when the leg affected by hemiparesis is in swing phase. During that period, the weight of the body is supported by the other leg (ipsilateral side). 20

35 Figure 12: With one leg in swing, a moment can be applied onto the pelvis about the weight supporting hip joint with just one actuator. Adapted from [9]. The center of rotation of the pelvis shifts with respect to the center of mass of the body throughout the gait cycle. Despite this, a single force with a carefully chosen line of action can exert a fully controllable moment onto the pelvis in the frontal plane. Here, the moment arm consists of a line segment perpendicular to the line of action of the applied force, and spanning between it and the hip joint of the supporting leg (this does not hold true during double support stance). This is illustrated in Figure 12 below. The RGR Trainer applies a corrective moment onto the pelvis only when the hemiparetic leg (or that assumed to be hemiparetic in case of healthy subject tests) is in swing. The RGR Trainer uses a synchronization algorithm, which produces an estimate of the subject s location in their own gait cycle, as explained in section 1.26, and the controller activates the force field only when the leg on the affected side is believed to be in swing. This 21

36 makes it possible to use only one actuator to generate a well-defined moment around the pelvis in the frontal plane, with a vertical reaction force at the support leg, which is equal in magnitude to the applied force generated by the actuator RGR Trainer Mechanical System Overview Figure 13: Subject in the RGR Trainer with major components labeled. The RGR Trainer controls one degree of freedom in the motion of the pelvis: obliquity. The remaining two rotational DOFs (pelvic rotation and pelvic tilt) and three translational DOFs are non-actuated (except for the ground reaction force on the foot of the non- 22

37 actuated side). The two major mechanical subsystems of the RGR Trainer, as shown in Figure 13 are: 1. Actuation system, which follows the natural motions of the subject s pelvis, while applying corrective moments to pelvic obliquity as determined by the control system. 2. Human-robot interface (HRI), a lower body exoskeleton, which is designed to transfer corrective moments to the pelvis. The HRI employs the waist, thighs, shanks and feet to effectively and reliably impart significant forces onto the user s lower body, and alter the orientation of the pelvis in the frontal plane (pelvic obliquity) Actuation System Force generation is achieved via the servo-tube actuator (model STA2508) from Copley Controls Inc. (Canton, MA, USA) a direct-drive electromagnetic linear motor, with windings in the actuator housing, and permanent rare-earth magnets in the movable thrust-rod (Figure 14). The servo-tube is a very good source of force and lends itself very well to impedance control [26]. Its total mass is 3kg, while the inertia of the moving thrust-rod is 1.9 Kg, with a 25cm stroke. This actuator can output 102N continuously, and up to 624N peak force (for 1 second). In some applications like mobile exoskeleton actuation, this linear motor s force density (ratio of force output to mass) is sub-optimal, but the RGR Trainer mitigates this shortcoming by supporting the housing of the actuator. The subject does however experience inertia of the housing and the moving thrust-rod in horizontal motions (translations and rotations). The thrust rod is extended 23

38 by a precision shaft, which is guided by two linear ball bearings. A spherical joint is used to transfer forces from the actuator to the brace, while a tension-compression load cell provides force feedback for control and performance evaluation purposes. Hall-effect sensors provide actuator position feedback by sensing the series of permanent magnets in the thrust rod (Figure 14). Figure 14: Servo tube linear actuator from Copley Controls Inc. 24

39 A lightweight assembly with a linear potentiometer provides vertical position feedback on the side opposite of the actuator, as shown in Figure 15. The actuation system of the RGR Trainer is suspended over the treadmill with a Biodex II frame (Biodex Medical Systems Inc.), and two sets of linear guides on each side (Figure 13) or a linear guide and a rotary joint on each side (Figure 16) depending on the development stage of the RGR Trainer. Both designs enable the actuation system to follow the subject in the horizontal plane with little friction, while resisting forces in the vertical direction, while the newer design (Figure 16, 17, 18, 19) shifts the vertical component of the Biodex II frame back behind the subject, giving easy access to the subject s legs while in the device. All tests presented in 0 were performed with the version of the RGR Trainer horizontal motion system pictured in Figure

40 Figure 15: Sensing (pelvis orientation and force) and actuation in the RGR Trainer. The servo-tube actuator (which contains hall-effect sensor based internal position measurement) and the linear potentiometer are fixed to the frame of the RGR Trainer, but can follow the motion of the body in the horizontal plane. 26

41 Figure 16: Horizontal motion system upgrade with the actuation system. Triangular subassemblies support the linear actuator assembly and the linear potentiometer assembly. A revolute joint about the vertical axis and a prismatic joint in the horizontal plane provide unconstrained motion in the horizontal plane while constraining motion in the vertical direction. 27

42 Figure 17: The horizontal motion system (upgrade) with the actuation system is secured to the Biodex frame with four locking clamp subassemblies. Body weight support components are not shown. Each structure is supported with two tapered roller bearings, which themselves are located concentrically with precision shafts. The mounting block shown in Figure 18 is bolted directly to the Biodex frame upright. 28

43 Figure 18: Revolute joint detail showing one tapered roller bearing mounted on a precision shaft. Two opposing bearings support axial loads in either direction, and radial loads. Set screws are used to lock the mounting blocks to the precision shaft and keep the distance between the bearings fixed. Figure 19: Complete horizontal motion system ready to be mounted on the Biodex frame. 29

44 1.15 Human - Robot Interface Initially an off the shelf Newport 4 pelvic brace (Orthomerica Inc.) designed for post-op hip revision was used to transfer forces to the subject s lower body, as shown in Figure 20. Experiments with the RGR Trainer have shown that using the pelvic brace alone (without the thigh segments) results in significant migration on the body. Adding the thigh segments (see Figure 20) improved force transfer capabilities, but it caused interference issues between the legs and made walking uncomfortable, restricting hip abduction and still not stopping migration of the brace on the body. Therefore, it was decided to design and build a new pelvic brace system, which would allow the user to retain natural gait pattern while enabling transfer of high forces to the pelvis in the frontal plane, to control pelvic obliquity. Figure 20: Newport 4 pelvic brace with thigh segments attached. 30

45 Lower Body Exoskeletons with Design Features Relevant to Pelvic Motion Lokomat The standard Lokomat device mentioned in section 1.9 has 4 active DOFs (2 knees and 2 hips) as well as body weight support system and vertical displacement. Researchers at the University Hospital Balgrist in Zurich suggested in that such a design, which restricts lower limb motion to just the sagittal plane may have negative effects on neural recovery [27]. This is due to reduced shifting of body weight between the two legs and insufficient excitation of the cutaneous, muscular and joint receptors. Therefore, the standard Lokomat mechanized orthosis was modified to include three additional actuated DOFs: lateral pelvic displacement and left/right hip abduction/adduction, and a nonactuated vertical displacement DOF was outfitted with an actuator to compensate for disturbing inertial forces. Subjective assessment of several healthy subjects tested in the modified Lokomat was that inclusion of the additional DOFs made training more physiological. Addition of these extra DOFs in the modified Lokomat orthosis still does not explicitly allow either pelvic obliquity or pelvic rotation, because the relative position of the two hip joints (relative to each other) is still fixed LOPES The LOPES exoskeleton shown in Figure 21 has 8 actuated DOFs (knee and hip flexion/extension, hip abduction/adduction, and lateral and forward-back pelvis translation) and 1 non-actuated DOF (vertical pelvis displacement) [19]. The LOPES team conducted a study which investigated the effect of pelvis fixation of gait characteristics [17]. The study concluded that fixation of the horizontal motions 31

46 (translations) of the pelvis during treadmill walking significantly changed almost all gait descriptors. The LOPES does allow all three translations of the pelvis, but pelvic obliquity and pelvic rotation are not explicitly allowed. Figure 21: LOPES DOFs, including free hip flexion/extension and free abduction/adduction BLEEX The BLEEX (Figure 22) is a mobile exoskeleton which features 7 DOFs per leg (4 actuated and 3 un-actuated) [28]. The actuated DOFs are: hip flexion/extension, hip abduction/adduction, knee flexion and ankle plantarflexion/dorsiflexion. The first two DOFs are co-located with those of the user thanks to an intricate pelvic brace design, shown in Figure 23. The BLEEX s un-actuated DOFs are: hip rotation, ankle abduction/adduction, ankle rotation and toe flexion/extension. Since this exoskeleton is supported by an external frame, and with hip abduction/adduction allowed (and actuated), BLEEX does allow pelvic obliquity, while un-actuated hip rotation allows for pelvic 32

47 rotation. therapies. Nevertheless, this device was never intended to administer gait retraining Rather, the BLEEX is an energetically autonomous lower extremity exoskeleton capable of carrying payloads. Figure 22: BLEEX worn by a user. Figure 23: Pelvic brace design of BLEEX exoskeleton. 33

48 RGR Trainer s Human-Robot Interface Design General Description Figure 24: Complete human-robot interface suspended from the RGR Trainer s actuation system front view. Our human-robot interface (HRI), shown in Figure 24, is composed of three major subassemblies: the pelvic brace and two leg braces. This system was designed to maximize effectiveness of force transfer to the pelvis, while minimizing time and effort necessary to don and doff the system. Therefore, this exoskeleton spans all the major joints in the lower body: ankle, knee and hip. By linking the exoskeleton s ankle braces with the pelvic brace using rotational joints which are almost all co-located with those of the user, it becomes possible to employ the majority of the lower body to transfer 34

49 moments to the pelvis. Also, strapping the human-robot interface to the relatively dense feet solved the migration issue characteristic of the earlier method (Figure 20). Two plastic shells (Newport 4 brace) wrap around the pelvis and locate the HRI with respect to the body in the horizontal plane as shown in Figure 25. Figure 25: HRI top view. Plastic shells wrap around subject s pelvis Design Details The human-robot interface has various free DOFs and adjustments to lower body size and shape, as shown in Figure 26. Each half of the HRI explicitly accommodates 5 DOFs. They are: - hip abduction/adduction - hip flexion/extension - hip internal/external rotation - knee flexion/extension - ankle plantarflexion/dorsiflexion 35

50 Ankle inversion/eversion is accommodated implicitly through shifting and play in the fit of the ankle brace inside the shoe. Through proper HRI adjustment to the subject, all the DOFs can nearly coincide with the subject s joint axes, except for hip internal/external rotation axes, which are shifted several inches away from the anatomical axes. Figure 26: Right side of the human-robot interface, with all DOFs (left) and adjustments (right) shown. Plastic shell interfacing with subject s waist was removed for clarity. The DOF axes are: (1) hip flexion, (2) hip abduction, (3) hip internal/external rotation, (4) knee flexion, (5) ankle flexion. Adjustments: (a) hip joint span, (b) pelvis width, (c) thigh length, (d) shank length, (e) knee frontal plane angle. 36

51 The human-robot interface was designed to fit the U.S. population ranging between 1 st and 99 th percentile (men and women) [29]. In order to accommodate a variety of body sizes and shapes, the HRI has the following adjustments: - hip joint span - pelvis width - thigh length - frontal plane knee angle - shank length Four identical rotational joints (Figure 27), consisting of two roller bearings and thrust bearings each, link the rigid aluminum structure around the pelvis to the leg braces. Figure 27: Hip revolute joint, with potentiometer for flexion-extension angle measurement. Precision shaft is double supported by a pair of needle pin bearings and thrust bearings. 37

52 The knee joints feature adjustment of the frontal plane knee angle. Including this feature in a compact design with space for knee flexion/extension measuring potentiometer was quite challenging. Our knee joint design is shown in Figure 28. The potentiometer s rotor is aligned with the centers of rotations of the two spherical joints, and rotates with the shank component (held with set-screw), while the body of the potentiometer rotates with the thigh component thanks to a music wire spring. Figure 28: Knee joints with adjustable frontal plane angle (two extremes shown) and rotary potentiometer for knee flexion/extension measurement. The design is optimized to resist moments in the frontal plane, resulting from force fields applied to pelvic obliquity Manufacturing The load carrying components of the human interface system were designed to withstand forces resulting from the structure supporting full weight of a 244Lb (111 kg) subject (99 th percentile male [30]), with a safety factor of 2. The major components of 38

53 the pelvic brace were machined from high strength aluminum alloy The components which are subject to low forces were rapidly prototyped using an in-house stereolithography (SLA) machine Conclusion The final design of the RGR Trainer is a product of multiple revisions and updates, which were motivated by continuous testing. The result is a practical, robust and reliable design, which allows the user to ambulate with little restraint to all the natural lower body joint articulations, while enabling transfer of high forces to the pelvis in the frontal plane via the feet, shanks, and thighs. 39

54 RGR Trainer Control System 1.17 Introduction The following sections outline various components of the control system which the RGR Trainer employs to accurately and intelligently administer gait-retraining force fields to the user Impedance Control Theory Here, when speaking of impedance control, we refer to the control of the end-point impedance of a robot or an actuator. Impedance control architecture consists of an inner unity feedback force loop, and an outer unity feedback position loop. The main task of the force loop is to increase backdrivability of the actuator. In that sense, force feedback moves any actuator closer to an ideal source of force. The outer position loop sets the relationship between the position of the end-effector, and the force it exerts. This is usually accomplished with a PD controller, where the proportional term represents virtual spring stiffness, and the derivative term acts like a virtual damper. A simple schematic of an impedance controller is shown in Figure

55 Figure 29: Basic outline of impedance control architecture. Proportional and derivative gains (PD) produce a force command which is executed by the force loop with gain G. System s interaction force with the environment (F ext ) is measured with load cell Effect of Force Feedback on Actuator Mass The analysis presented in here is based on Hogan s work in [31], which was adapted for control of a linear actuator. Neglecting friction, the actuator s thrust rod can be represented as a mass m undergoing displacement x due to forces F act applied by the actuator s electromagnetic field, and F ext, or external force, applied by the environment. mact x Fact Fext (4.1) Fext mact Fact X 41

56 Figure 30: Simple model of actuator s thrust rod. The equation describing a simple closed loop control law is: F G( F F ) (4.2) act ref ext These two equations combined give us the following equation: And the transfer function is: m x G( F F ) F (4.3) act ref ext ext X GF ( G 1) F ref ext (4.4) 2 mact s This can be represented by the block diagram below: F ref G 1/m act dx 2 /dt F ext Figure 31: Actuator shaft and force control law. The immovable mass (body) with stiffness and damping, with F ext being the interaction force between the body and the actuator, can be represented by the first order equation: and its Laplace is: Fext Be x Kex (4.5) X Fext B s K e e (4.6) 42

57 Now we equate the actuator transfer function (Equation 4.4) with the body s transfer function (Equation 4.6) to describe the actuator body interaction: F F ext ref G ( Bes Ke) ( G 1) mact ( G 1) 2 s Bes Ke (4.7) Where m act /(G+1) is the apparent inertia as experienced by the environment. Therefore, the effect of force feedback is the reduction of the apparent actuator inertia by a factor of G Derivation of Impedance Controller In [31], Hogan presents an impedance controller for stable contact execution between a robot and the environment. The following impedance controller derivation is an adaptation of Hogan s work for controlling actuator s end point impedance in the RGR Trainer. The simplified actuator dynamics are shown in the figure below. Fext mact Fact X Figure 32: Dynamics of the actuator. 43

58 The equation describing the dynamics is: mact x Fact Fext (4.8) where the force generated by the actuator (F act ) onto the thrust rod is: Fact mact x Fext (4.9) The desired end-point impedance of the actuator thrust rod can be represented by the following equation: F M ( x) B ( x x) K ( x x) (4.10) ext c c o c o where M c is the actuator s apparent mass (inertia), B c is controller derivative gain (damping) and K c is controller proportional gain (stiffness). The desired acceleration of the actuator thrust rod is: 1 x [ K ( x x ) B c ( x x ) F M c c 0 o ext (4.11) Now substitute the desired acceleration into the actuator force equation: and m F [ K ( x x) B ( x x) F ] F (4.12) act act c 0 c o ext ext M c m mact F [ K ( x x) B ( x x)] F [1 ] M act act c 0 c o ext Mc c (4.13) 44

59 Equation (4.13) above describes the impedance controller. F act is the force command sent to the servo-amplifier. We would like the inertia of the thrust rod mass - m act, to be as low as possible. In practice, the degree to which this apparent inertia can be reduced by use of force feedback is limited. We equate the desired mass M c to the lowest possible apparent inertia of the thrust rod: M c = m act /(G+1) and the force controller gain G is picked to be highest possible, while still providing appropriate stability margin. After the substitution, the equation describing force commanded to the actuator F act is: F ( G 1)[ K ( x x) B ( x x)] ( G) F (4.14) act c 0 c o ext The above equation lists the constituents of the force command F act, which is sent into the servo amplifier, to be executed by the actuator. This can be represented by the following diagram: Reference Trajectory + - Position Error K c + B cs F virt G+1 F ref + - F act Amplifier & Actuator Body F ext Position Feedback G Force Feedback Figure 33: Physical implementation of equation (4.14). The output of the PD controller, which acts on the position error, can be called the virtual force, F virt. It is the output of the virtual spring and virtual damper, K c and B c respectively. It is a known fact that force controller gains are often limited to single digits. At such low gain values, the steady state error can be very significant. For example, using the 45

60 control law of equation (4.2) and a proportional gain G=1, the resulting force output F ext is only 50% of the reference F ref. The impedance controller from Figure 33 takes this effect into account, magnifying the PD controller s output by (G+1) to cancel the following-error resulting from the control law and low gain value. Due to force feedback s dependence on the environment, tuning is often performed manually [26] Actuator Position Feedback The servo tube actuator is equipped with hall-effect sensors, which are used by the Xenus servo amplifier to generate an emulated differential quadrature encoder signal (position). The differential encoder position signal from Xenus is converted to single ended using a US Digital Inc. incremental encoder adapter. The encoder signal is acquired by the National Instruments Inc. 6259M DAQ card, (80MHz hardware counter), counting both rising and falling edges of the incremental encoder signal (X4 encoding). The net number of counted edges is polled by the controller at 500Hz and converted to position with knowledge of encoder s resolution (12.5 microns). The linear potentiometer s signal is low-pass anti-alias filtered (RC 480Hz cutoff), and acquired by the DAQ at 2kHz. Pelvic obliquity angle is computed as shown in Figure 52, at the control loop s operating rate (500Hz) Force Feedback The degree to which the actuator system can actually display the specified endpoint impedances depends largely on the extent of backdrivability of the actuator. The higher the backdrivability, the better the system can display the commanded forces. Therefore, 46

61 proper implementation of force feedback is crucial to implementation of impedance control. The signal from the Honeywell-Sensotec model 31 compression tension load cell is amplified by the Honeywell-Sensotec UV-10 in-line amplifier. A sample of raw load cell data is shown in Figure 34 below. An analog anti aliasing low pass RC filter with a cutoff frequency set to 480Hz greatly improves the signal quality, as can be seen in in Figure 35. Figure 34: The unconditioned load cell signal contains significant noise. 47

62 Figure 35: Once analog filtered, the noise level in the signal is greatly reduced. Figure 36: Unfiltered load cell signal sinusoidal loading at 1Hz. 48

63 Figure 37: Low pass analog filtered (480Hz cutoff) load cell signal sinusoidal loading at 1Hz. As can be seen the power spectrum density plots in Figure 36 and 37, much of the noise which existed in the unfiltered signal in the 0 to 480Hz range was largely attenuated by the low pass filter. This suggests that these attenuated signal components were aliases of higher frequency noise (above 480Hz). A 4th order inverse Chebyshev filter, with 30Hz cutoff and a 60dB attenuation level conditions the signal further, as pictured below. 49

64 Figure 38: Combination of analog filtering to remove high frequency aliases, and digital filtering produces a clean force signal Control Hardware and Software The control hardware consists of a 6259M PCI data acquisition card (DAQ) from National Instruments (Austin, TX, USA), which runs in LabVIEW Real-time operating system (RTOS) on a dedicated PC (target) and a 4-core CPU. A host computer (laptop) serves as the user interface. The servo tube linear actuator requires the Xenus servo amplifier for operation. This digital amplifier can be programmed to operate in three different modes: position, velocity or force. 50

65 Figure 39: Xenus servo amplifier from Copley Controls Inc. In the force mode, the Xenus servo amplifier does not use a direct measure of force, but it does however monitor the current consumed by the actuator, and a proportional integral (PI) controller adjusts the voltage sent to the actuator in order to coax the requested current draw. The block diagram of the amplifier s internal control loop is pictured below. An automatic tuning procedure performed by the Xenus servo amplifier set the current loop gains to C p = 454 and C i = 88. Figure 40: Outline of the inner current loop contained in the Xenus servo amplifier. 51

66 1.22 Force Controller Tuning Methods and Materials The actuation system was first bench-tested in open loop mode, followed by closed-loop mode. A healthy subject wore the Newport 4 pelvic brace with a thigh segment on the left leg. The actuator body was fixed to a vertical rail, while the actuator s thruster was connected to the pelvic brace through a compression tension load cell and a spherical joint as shown in Figure 41. The performance of closed loop force control depends largely on the mechanical properties of the environment, gains are often found by manually tuning the force controller [26]. Therefore, we explored a range of force loop gains, while the control law from Equation (4.2) was implemented Open Loop Test Protocol The subject remained still, while a step input of 50N (downward force) was sent to the amplifier using LabVIEW, and the resulting force between the actuator shaft and the pelvic brace was measured and recorded at 200Hz. 52

67 Figure 41: Actuator shaft coupled to the body via pelvic brace, with load cell reading the interaction forces Closed Loop Test Protocol The proportional gain G was varied between 0.6 and 1.8 as the step response tests were performed (50N downward force). 53

68 Results The response of the open loop system can be seen in Figure 42 below. Despite the fact that the servo amplifier tracks current consumption with its internal closed loop controller, there is significant steady state error in the system s output. Figure 42: In open loop mode, we see about 15-20% steady state error. Plots of closed loop force controller response at two gain settings, G=1 and G=1.8, are shown in Figure 43 and

69 Figure 43: Closed loop step response with proportional gain G=1. Figure 44: Closed loop step response. With a proportional gain of 1.8, serious instability occurred. 55

70 Discussion It is interesting to note that the gain of G=1 results in a steady state error of approx. 50% (Figure 43) and is a direct result of the control law and the low gain used. Vibrations as perceived by the subject wearing the pelvic brace became significant once the gain G=1 was used, and an instability of the system occurred at G=1.8, as can be seen in Figure 44. It has been shown in [32] that in force control, a proportional force loop gain of 1 or larger can cause instability. When the system operated in backdrivable mode, during which the control system actively minimizes the interaction forces, we found that setting G = 1offered best performance (highest interaction force reduction) with no adverse effects (perceptible vibrations). When that same controller was used to generate force fields, we found that force gain reduction from G=1 to G=0.7 was necessary to reduce the perceptibility of vibrations which became more noticeable in that operating mode. As can be seen from the step input graph, gain G=1 produces a steady state error of 50% in the output with respect to the reference input. That result alone is unacceptable. Fortunately, this effect is compensated for by the impedance control algorithm. In Equation (4.14) which was derived earlier for force command sent to the actuator, the sum of the two products in the square brackets amounts to F virt the virtual force necessary to generate the desired force-field, which should be generated by the system. This virtual force is multiplied by the factor (G+1). The product is the reference force, F ref, the magnitude of which is amplified ensure that the force commanded to the actuator matches the force specified by the impedance gains (F virt ). Therefore, the steady state and tracking error observed during force loop tuning is eliminated. 56

71 1.23 Linear Motion Impedance Controller Bench Tests After tuning the force loop, another set of tests was performed to characterize the end point impedance controller. The main purpose here was to find its operating envelope. In our application, the core function of the impedance controller is to coax the actuator end-point to display virtual stiffness and damping. In reality, as we have seen, the actuator also has inertia. Therefore, these tests also revealed the effects of the apparent inertia of the actuator, along with any friction and noise inherent in the system while it attempted to display the aforementioned virtual spring and damper qualities Methods and Materials The end-point impedance controller, as described by equation (4.14) was first realized for control of linear motion of the servo tube actuator (Figure 41). This simplified the benchtesting of the system. The findings from these tests were fully applicable to the pelvic obliquity (rotational motion) impedance controller, which was constructed subsequently. Figure 45: Linear motion impedance controller was used in bench testing. Since our impedance controller has at its core a force controller, its performance depends largely on the dynamic properties of environment which it interacts with. Therefore, for 57

72 these tests, the actuator s end-point was attached to the pelvic brace worn by a subject just like in the force tuning experiment (Figure 41). There are three possible inputs into our impedance controller: reference trajectory (position), measured position and measured force. Since it would have been difficult for the subject to produce repeatable position displacement, the subject was instructed to keep his body still while sinusoidal reference trajectories of different frequencies were presented to the controller. One goal of the test was to find how well the actuator displays the commanded dynamic behavior. Therefore, the system s load cell force data (F ext ) was acquired and compared against the force commanded by the impedance controller (F virt ). All tests were conducted with the force loop proportional gain G=1. The physical setup from the force loop tuning tests was used again (Figure 41). For convenience, the derivative gain was specified through a selection of damping ratio (zeta ζ). The equivalent derivative constant B c was calculated from the knowledge of actuator s moving mass, specified stiffness K c and desired damping ratio. The actuator itself was represented as a second order system. The characteristic equation is: s 2 Bc b Kc ( ) s (4.15) m m m where b is the damping inherent to the actuator. The standard representation of second order characteristic equation is: s 2 s 0 (4.16) 2 2 n n One can compare the two equations above and arrive at the following result: 58

73 B c 2 m* K (4.17) c Protocol A sinusoidal cycle (3cm amplitude) served as the position reference trajectory, and the response of the system (position and force) was acquired and saved at 500Hz. Tests were conducted at three different input frequencies (1, 3 and 6Hz), three different proportional gain values: K c =1, 5 and 10kN/m, and within these, a range of damping ratios was also tried, ranging between 0 and Results Due to the volume of data collected and analyzed, only the most significant results are shown in this section, while the remaining results are located in Appendix A. The figures below confirmed that in the general sense the impedance controller generated a proper magnitude force field. As the position error increased, so did the forces commanded by the impedance controller. The offset between the virtual force F virt and the measured force F ext is due to the offset between the desired position and the actual position. The interaction force between the brace and the actuator somewhat closely followed the virtual force F virt, with some oscillatory behavior (vibrations) when the damping ratio was set to zero, as we can see from Figure 46 and 48. Introduction of damping attenuated these vibrations (Figure 47 & 49). 59

74 Figure 46: Position error (Des. Position Act. Position) generated the force command (F virt ). Load cell measured actual interaction force (F ext ). Figure 47: The addition of damping attenuated the oscillatory force interaction. Effects of stiction can be seen just past maxima and minima. 60

75 Figure 48: Higher gain value caused greater environment deflection (Act. Position). Lack of damping resulted in oscillatory response. Figure 49: diminished. After the damping ratio (zeta) was introduced, the oscillatory behavior Increasing the virtual spring value produced an expected result: the actuator system produced greater environment deflections by exerting larger forces. As the damping ratio 61

76 was increased during the test, vibrations in position and force diminished. Next, the stiffness was increased to 10kN/m. Some of the test results are presented below. Figure 50: At this high virtual stiffness setting, slight vibrations were again felt, and can be seen in the F ext signal. At a relatively high stiffness setting of 10kN/m, the controller behaved well, and the output forces closely matched the controller - commanded forces, with slight oscillatory behavior, which can be again attributed to lack of damping. 62

77 Figure 51: Increasing the damping ratio to 0.8 amplified the vibrations. With the damping ratio past 0.6, the controller s output signal began displaying significant high-frequency vibrations. They were perceivable to the subject, but due to low-pass filtering, these vibrations are not visible in the F ext signal. Reducing the damping ratio to ζ = 0.3 resulted in the greatest reduction of vibrations in the system Discussion While at the lower K c gain values tested (1 and 5kN/m) increasing the damping ratio up to ζ=0.8 had the effect of only improving the performance, at K c =10kN/m the damping ratio of ζ=0.3 resulted in optimum performance (lowest level of vibrations as perceived by the subject), and anything higher than ζ=0.5 caused significant deterioration in performance. This suggests that either the noise or the time delay (phase shift due to filtering) in the derivative signal is responsible for these vibrations. The same effect has been described in literature [25]. Therefore, a decision was made to establish a limit on the damping ratio and position and force proportional gain, below which these 63

78 undesirable vibrations do not occur. In light of the results from the preceding tests, the operating envelope of the system was defined by limiting the maximum allowable damping ratio to ζ =0.5, and the virtual stiffness K c to 10kN/m. The set of tests conducted made it possible to move forward with developing a pelvic obliquity impedance controller Pelvic Obliquity Position Feedback As we saw in section 1.12, the RGR Trainer applies a moment to the pelvis in the frontal plane, to affect the pelvic obliquity angle. This task requires measurement of the pelvic obliquity angle at all times throughout the gait cycle, as well as measurement of the moment or force exerted onto the subject by the RGR Trainer. In the field of motion analysis, pelvic obliquity is specified in degrees of angular rotation. In order to comply with this standard, we decided to offer position feedback to the controller in the same format. The RGR Trainer uses two linear position measurement units, which are attached to either side of the pelvic brace and operate in the vertical direction. These units are a linear potentiometer and an emulated encoder (internal to the actuator). Position feedback coming from the linear actuator is described in section The pelvic obliquity angle of the pelvic brace is calculated using the relative position of the two attachment points on the pelvic brace (in the vertical direction) and the distance between these two points. This is illustrated in Figure

79 Figure 52: Obliquity angle can be calculated knowing vertical position of the two attachment points. D is the length of the direct line between the two attachment points, and y is the distance between them in the vertical direction. As one side of the pelvis moves upward with respect to the other, the segment of length D spanning the two attachment points rotates. If this segment D serves as the hypotenuse of a right triangle and the difference in height between the two attachment points is y, then the resulting angle of rotation Θ is the pelvic obliquity angle Pelvic Obliquity Impedance Controller In order to apply impedance control at the obliquity level, the control algorithm from the linear motion case was adapted to act on angular position error measured in degrees of the pelvic obliquity angle. This system s block diagram is presented in Figure 53, and the details of the PD controller block are shown in Figure 54. The strength of the force field is specified with the proportional gain K c, with units of N-m/deg. For convenience, 65

80 the derivative gain B c is not specified independently, but is computed based on the damping ratio using equation (4.17). This type of approach to gain selection allows fast changes to be made to the force - field strength while the general dynamic properties of the system remain unchanged. The PD gains produce a force command, which is executed by the impedance controller s force control loop. Θ Reference Obliquity + - PD G Amplifier Actuator F Subject Θ Obliquity Low Pass Filter 2 (analog) G Low Pass Filter 1 (analog) Figure 53: The PD controller acts on the obliquity error and outputs the appropriate force command. Low pass filters 1 and 2 are RC anti-alias filters. Angular Position Error [deg] Kc [N-m/deg] (rotational) Moment Arm Force Command [N] d/dt Low Pass Filter 3 (digital) Angular Velocity Error Jacobian Linear Velocity Error Bc [N/m/s] (linear) Figure 54: Details of the PD gain block from Figure 53. The proportional gain K c is specified at the obliquity level, while the derivative gain B c acts on linear velocity error at the actuator level. B c is computed from K c (linear motion equivalent) and the specified damping ratio ζ using Equation Velocity feedback undergoes secondary filtering (after velocity error is computed). 66

81 1.26 Human Machine Synchronization Introduction Pelvic obliquity reference trajectory is a time series, containing the relationship between space and time. Therefore, in addition to being properly positioned in space, the individual data points of the reference trajectory also have to be presented to the impedance controller at the right time. Preliminary force field application tests with healthy subjects required that the subject actively adjust their temporal position and velocity (cadence) to match that of the reference trajectory, which was replayed at a constant speed. This took a considerable amount of practice and effort, and cannot be expected of an impaired subject. Therefore, a synchronization algorithm presented by Aoyagi in [6] was implemented in the RGR Trainer Synchronization Algorithm The duration of a single gait cycle spans between two consecutive left heel strikes. The right heel strike occurs at the 50% mark in the gait cycle (assuming symmetrical gait). The synchronization algorithm estimates the actual temporal position of the subject within his gait cycle based on the angular positions and velocities of the subject s hip and knee joints (8 degrees of freedom). A reference for the synchronization algorithm is constructed by recording an 8-dimensional time series over several gait cycles and finding the normalized mean of each DOF. The 8 DOFs are normalized to ensure that they are assigned equal weight. The reference is generated by the norms of the individual vectors, and is represented by the loop of discrete points in Figure 55. The number of discrete points in the reference is a function of walking cadence and sampling rate used. 67

82 Figure 55: Conceptual diagram and synchronization algorithm diagram, adapted from [6]. During operation, a minimization operation of the norm of the difference between the measured 8-dimensional vector and every vector in the reference is performed, and this identifies the location of the nearest neighbor. This result is normalized to give an index value ranging between 0 and 1. This represents the location of the subject in the temporal sense in the gait cycle Physical Implementation The human robot interface features knee and hip angle measurement (4 DOFs), as shown in section Taking derivatives of these signals produces four angular velocities, for a total of 8 DOFs for use in the Aoyagi synchronization algorithm. In addition, a low profile assembly with a micro switch, which is placed in the subject s left shoe, is used to detect left heel strikes. It is based on a mini push button switch COM ( The assembly shown in Figure 56 consists of thick aluminum sheet and 0.9 x0.9 PCB mounted with four 4-40 screws. Knowledge of such 68

83 a discrete gait event is useful for both generating synchronization reference trajectories, and for synchronization algorithm performance validation purposes. Figure 56: Foot switch construction. Clear plastic sheet taped over the top improves user comfort. The Aoyagi gait estimation algorithm was implemented as it is shown in Figure 55 in LabVIEW Real Time at 500Hz, on a dedicated CPU core (one of four available) due to its high computational load. Signals from the 4 rotary potentiometers (hip and knee joints) were analog low pass RC filtered, acquired by DAQ at 2 khz and digitally filtered. Heel strike signal, which is also collected, is used to parse the data and find 8 means of the 8 DOFs (hip and knee angular positions and velocities) across the multiple gait cycles. 69

84 Quantitative Evaluation of Performance One way to quantify the synchronization algorithm s performance is to compare its output to a nominal metric. While walking at constant speed and cadence with symmetric gait, a healthy subject s progression through his gait cycle generates a straight line, when plotted against time. We can create a useful metric by generating a linear interpolation between a cyclic discrete gait cycle event, such as left heel strike. This allows us quantify the performance of the synchronization algorithm. Therefore, a nominal time series was created by concatenating vectors generated by linear interpolation between the consecutive heel strikes in a sample dataset. Figure 57: MATLAB s unwrap function produces continuous curves of periodic time series. The range of phase angle of 5 radians to 75 radians covers approx. 11 full gait cycles. 70

85 This nominal position in the gait cycle is graphed in Figure 57 together with the unwrapped, filtered output of the gait estimation algorithm (T rep ). Low-pass filtering contained inside the gait estimation algorithm produces a time delay. An offset equal to the delay (0.53s) found in Figure 57 was applied to the data, nearly superimposing the two curves (Figure 58), with MSE = , or 6.98 % of the gait cycle. Figure 58: With offset introduced to remove delay, heel strikes as predicted by gait estimation algorithm ( T rep ) nearly coincide with those produced by the discrete gait event (heel strike) Overall Control System Architecture The complete control system shown in Figure 59 is built up around the pelvic obliquity impedance controller. This control system allows for modulation and fine-tuning of the force field applied onto the subject in two major ways. 71

86 Firstly, the controller can switch between two (or more if necessary) different position references while in operation, within two consecutive gait cycles. Subject s hip and knee joint angular positions and velocities are used by the Aoyagi gait estimation algorithm (section 1.26) to produce an estimate of the subject s point in the gait cycle at any time. This estimation of the point in gait is used in two lookup tables to generate two position references. Switch 1 shown in Figure 59 executes a transition between the two reference trajectories. This switch follows a sigmoid curve, which is a section of a 3Hz sinusoid, spanning between 0 and 1. Switch 1 is set to go on or off beginning at 20% of the gait cycle, when the contralateral leg is in stance. Figure 59: Overall Control System Architecture. The second way to control the force field applied onto the subject is through precise activation and de-activation of the impedance gains. Switch 2 in Figure 59 follows a sigmoid curve as well, enabling a smooth transition from the backdrivable mode (zero force control) to impedance control mode, when the PD gains set the desired stiffness and damping (the force field). Sample data in Figure 60 shows the operation of Switch 2. 72

87 The ability to precisely control the timing of force field activation within the gait cycle only when the contralateral leg (the leg on the hemiparetic side of the body due to stroke) is in swing, means that the moments applied onto the pelvis are not indeterminate, despite the fact that only one actuator is used to apply an external force, as shown in Figure 12. This allows for adjustments in the PD gains when the force field is in the de-activated state. Figure 60: Two consecutive gait cycles. Synchronization algorithm output predicts left heel strikes well, and gives good estimate of gait cycle location mid-stride. Gait estimation (Synchr Output) is the progression through the gait cycle from 0 to 1 (100%). Force field activation sigmoid switch (3Hz) was set to go on at 44% and off at 76%. Heel strike is marked by the rising edge of the Heel Strike Switch signal. 73

88 The layout of the hardware components of the RGR Trainer s control system is shown in Figure 61 below. The impedance controller runs in LabVIEW RT OS environment on a dedicated PC (Target), while the user interface is rendered by another PC (Host) with Windows OS. The Real-Time environment allows for controller operation which is never interrupted by non-critical tasks, as often happens in non-deterministic operating systems such as Windows. Figure 61: Layout of hardware components of the RGR Trainer s control system. 74

89 Interaction Force [N] 1.28 Actuation System Backdrivability Methods In order to confirm the effect of force feedback on backdrivability of our system, interaction force data was gathered from one subject, and analyzed. By setting the PD gains to zero, the impedance controller is reduced to a zero-force controller Protocol The subject ambulated at his comfortable walking speed (CWS=3km/h) in the RGR Trainer, while wearing the human-robot interface. Two force data samples lasting 200 gait cycles were collected (sampling rate of 50Hz was used) and saved under two force loop gains (G=0 and G=1) Backdrivability Test - Mean across 200 Strides G = 0 G = % Gait Cycle Figure 62: Interaction force data with the RGR Trainer s control system set to follow mode, under two force gain settings, collected at 50Hz. 75

90 Results Completing the two 200-stride trials took approximately 260 seconds each. While the maximum interaction forces ranged between +/-20N without force feedback (G=0), the range of interaction forces was reduced to approx. +/- 10N with gain G=1, as we can see in Figure Discussion In order to better understand the effect of force loop gain on the system s backdrivability, power spectral densities of the time series from Figure 62 were estimated, as shown in Figure 63. The bulk of the interaction forces are contained within the 0 to 7Hz range. Use of force feedback with proportional gain G=1 resulted in interaction force attenuation of about 50% (6dB). This is a direct result of the force control law and the gain used. Figure 63: Backdrivability test results (with force control gains as indicated). The healthy subject ambulated at his comfortable walking speed (CWS) of 3km/h. Approximately 50% (6dB) interaction force reduction for frequencies 0-6Hz can be seen. 76

91 1.29 Actuation System Bandwidth We demonstrate in the previous section, that the bulk of low-magnitude force interactions between the subject and the actuation system occur within the range of 0 to 7Hz. The actuator s force bandwidth was characterized in order to ensure that it is sufficiently high for the task. Figure 64: Bandwidth test setup with two compression springs (k=5.66kn/m) Methods The servo tube actuator s force bandwidth was measured at two amplitudes: 10N and 60N. For safety reasons two compression springs (5.66kN/m stiffness) were used in place of a subject (see Figure 64). The stiffness of the spring was chosen based on the results of tests in section 1.22, where the displacement and force plots (versus time) 77

92 Force [N] suggest that the pelvic brace used for the bench-tests displayed a stiffness of approx. 5.4kN/m. Lack of damping in the springs (versus damping inherent to a subject) made it impossible to operate the system in closed-loop mode. Therefore only the open-loop bandwidth was tested. For each force level, a chirp signal (from 0 to 30Hz) was sent to the Xenus servo amplifier, and a load cell measured the interaction force (see Figure 65) N Force Bandwidth Test Chirp Force Command Interaction Force Frequency [Hz] Figure 65: Commanded force (Chirp Force Command) and the resulting interaction force Results Bode plots for the two force levels are shown in Figure 66. The open-loop bandwidth was found to be 14.7Hz for low forces (+/- 10N) and 12.1Hz for high forces (+/- 60N). During the 60N test, at the resonant frequency the interaction forces as measured by the load cell (Figure 64) reached approx. +/-400N. 78

93 Phase [degrees] Magnitude [db] Bode Plot N 60N Frequency [Hz] Frequency [Hz] Figure 66: Actuator force bandwidth test results Discussion The backdrivability test revealed that the bulk of the interaction forces occurring at the lower end of the operating force range is contained within 0-7Hz (Figure 63). The bandwidth test results (Figure 66) suggests that in open loop, the actuation system can in fact execute force commands at those levels up to 14.7Hz Safety The Xenus servo amplifier employs a Schmitt trigger in its enable function to recognize the enable signal ( greater than 3.65V ) and the disable signal ( less than 1.35 V ). In general it is desirable to ensure that the drive be disabled when the control software or the 79

94 computer itself fail. A simple solution of using a DAQ analog output channel to send an enable signal while the controller is in operation, and a disable signal when it stops is not perfectly reliable. Unfortunately, when LabVIEW software fails, the National Instruments 6259M card continues to output the last value for as long as there is power supplied to the computer. This creates a potentially dangerous situation, especially with a highly back-drivable system as ours working under impedance control. For example after a software failure, a high - force command from the time of failure will remain. Physically unplugging the servo amplifier s mains electrical supply would cause the actuator to cease exerting force thus becoming easily backdriven to a new position. Then, returning main power to the amplifier could cause a sudden, unexpected acceleration of the thrust rod, possibly causing injury and damage to the actuator assembly. To address this problem, one solution is to use an external microchip with a watchdog feature, which detects software failure, as was done in [20]. Our solution was to use a simple analog circuit to provide the enable signal to the Xenus servo amplifier only when LabVIEW is active. A dedicated DAQ output was configured to supply a sinusoidal voltage signal of 100Hz frequency and ranging between 0V and 10V. This signal was routed through an analog RC high-pass filter, with the cutoff frequency on the order of several Hz to avoid excessive signal attenuation. Then, the signal was rectified with a Gratz bridge rectifier and smoothed with help of a capacitor placed in parallel, as is shown in Figure 67. The result was a slightly varying voltage output which successfully enabled the Xenus servo amplifier when the input was of proper frequency and magnitude. At the same time, the circuit s output changed to 0V whenever the input was 80

95 0V or un-varying (as in the case of software error). Throughout bench testing, this scheme has been proven to work very well. Using an equation governing discharge of a capacitor: Vc t/ RC V0e (4.18) we can solve for time t: V t RC V 0 ln( ) (4.19) With R = 820kOhm and C1=C2 =4.7μF, the time for the voltage to drop from maximum 10V to Schmidt trigger s 3.65V on limit is 3.9s, and dissipation from 5V takes 1.2s. C Figure 67: Analog amplifier-enable safety circuit Conclusion The RGR Trainer s control system, comprised of the impedance controller and auxiliary components such as the synchronization and force field switching algorithms, makes it possible to accurately and intelligently apply corrective force fields to pelvic obliquity. 81

96 With just one actuator, such functionality would not have been possible without the precise timing of force field activation. The next chapter describes the various experiments, which we performed on healthy subjects. These experiments put the individual subsystems of the RGR Trainer to the test, proving the RGR Trainer s usefulness as a device for gait retraining. 82

97 Healthy Subject Testing 1.32 Introduction Gait rehabilitation techniques fall under two categories: neurophysiological or motor learning. In the classic neurophysiological gait rehabilitation techniques, the physical therapist (PT) supports the patient s correct movement patterns, acting as a problem solver and decision maker, rendering the patient a relatively passive recipient [33]. On the other hand, motor learning techniques stress active patient involvement in contextspecific motor tasks. Robotic gait rehabilitation is intensive, repetitive and task-oriented, and therefore aligns itself well with motor-learning techniques. Design of optimal gait retraining therapies for patients post stroke requires thorough understanding of normal, as well as pathological gait patterns. Human development period provides some insight into the former. Infants exhibit stepping behavior even before birth, thus before the motor cortex in the brain can effectively send motor commands to the muscles involved, but their stepping pattern lacks proper ankle joint control for ground clearance, resulting in toe-drag. Impaired dorsiflexion during the swing phase of the gait cycle is also the most common problem in walking after injury to motor areas of the brain [34]. Locomotion results from intricate dynamic interactions between a central program and feedback mechanisms [34]. The central program, contained in the spinal column s central pattern generators (CPGs), can generate basic locomotion patterns to steer, trigger and stop locomotion while feedback from muscles, skin afferents, vision, hearing, and 83

98 sense of balance are used to dynamically adjust the locomotion pattern to the requirements of the environment [35]. Therefore, it is the various brain regions such as the cerebral motor cortex, cerebellum, and brain stem, which are responsible for the fine control of the gait pattern. Recent work suggests that both peripheral sensory information and inputs from the motor cortex reshape the function of the CPG, especially during brain plasticity post-stroke. In fact, it is believed that the locomotor patterns generated by the CPGs are not sufficient for over-ground walking, and supraspinal control is needed to provide both the drive for locomotion as well as the coordination to negotiate a complex environment [36]. Motor adaptation studies, which were first conducted by Shadmehr and Mussa-Ivaldi [37] for upper extremities, are now being used on lower extremities, in order to better understand the mechanisms governing locomotion. Motor adaptation is the modification of a movement from trial-to-trial based on error feedback in which the following criteria are met. First, the movement retains its identity of being a specific action (e.g. reaching ) but changes in terms of one or more parameters (e.g. the pattern of force or direction). Second, the change occurs with repetition or practice of the behavior and is gradual over minutes to hours. Third, once adapted, individuals cannot retrieve the prior behavior; instead, they show after-effects and must de-adapt the behavior with practice in the same gradual, continuous manner back to the original state [38]. The reactive or feedback-driven adaptations differ importantly from predictive adaptations in that they occur more quickly in response to ongoing afferent feedback (i.e., do not require practice) and are not stored by the nervous system (i.e., do not produce aftereffects) [39]. 84

99 Since motor adaptation is one of the factors responsible for variability in rehabilitation outcomes following robot-assisted gait retraining, it is hoped that subjects performance in motor adaptation studies will be predictive of rehabilitation outcomes. Therefore, currently motor adaptations in lower extremities are being studied by several teams. A direct extension of the method first used by Shadmehr and Mussa-Ivaldi which applies velocity-dependent perturbations perpendicular to the direction of motion was implemented in the Lokomat by Cajigas et.al [40]. Motor adaptations were also investigated on a number of healthy subjects in the active leg exoskeleton (ALEX) [41]. In the LOPES [42], motor adaptations were studied while the goal of the training was to affect foot ground clearance during swing, by applying a vertical force proportional to the horizontal velocity of the ankle. Such choice of protocol was inspired by work done on upper limb motor adaptations, but interestingly some of the researchers findings contradicted their hypotheses. Therefore, they urge caution in implementing upper limp research principles for lower limbs. Another team, which used split belt treadmill gait training with stroke survivors was able to induce after-effects in gait symmetry [43]. Both this study and another one with upper extremities have found that adapting the patient to a perturbation that worsened or amplified their error was what drove adaptation to result in after-effects that improved their movement [38]. The after-effects were transient, and diminished in a matter of minutes, a process called wash-out. That s why researchers are investigating motor learning Motor learning is the formation of a new motor pattern that occurs via long-term practice (i.e. days, weeks, years). After the new movement pattern is learnt, it is stored 85

100 and can be immediately brought up and used in the appropriate context (i.e. in contrast to adaptation, no practice period is required). Individuals may store many learned motor plans or calibrations that allow for efficient switching from one to another [38]. The hope is that since clinicians often work with patients who can perform certain tasks such as walking or reaching, but their movements are slow, misdirected, inaccurate, or inefficient, it may be possible to help them recalibrate these movements, instead of learning completely new movement patterns. Hip-hiking, which is an abnormal pelvic obliquity pattern, is the most common secondary gait deviation, and while motor learning principles are being increasingly used to study locomotion, no such studies have ever targeted the control of pelvic obliquity. This motivated the various protocols described in this chapter, which aimed to: - Demonstrate that the RGR Trainer can effectively guide the pelvis in the frontal plane via force fields to alter pelvic obliquity. - Investigate whether the RGR Trainer can induce motor adaptations in pelvic obliquity control. The protocols presented here constitute the most significant approaches which were undertaken in healthy subject tests in the RGR Trainer. All subject tests were conducted in accordance with methods approved by the Spaulding Rehabilitation Hospital s Internal Review Board (IRB). 86

101 1.33 Protocol Hypothesis The RGR trainer can teach the hip-hiking gait pattern to healthy adults by generating an appropriate force field around their pelvis Protocol 1 Methods Experimental Apparatus and Reference Trajectories The RGR Trainer, configured to apply vertical forces on the left side of the body, was programmed to switch between two reference trajectories: baseline and hip-hiking. The switching action was designed to happen quickly but smoothly, occurring when left leg is in stance (due to the small position error at that time) and following a sigmoid curve at a frequency of 3Hz. The sigmoid is one half cycle of a 3Hz sinusoid, minimum to maximum amplitude or vice versa, spanning between 0 and 1. A hip-hiking pelvic obliquity reference trajectory was found in literature. A 2009 study by Cruz et al. [44] focused on understanding the potential causes of reduced gait speed and compensatory frontal plane kinematics during walking in individuals post-stroke. The data gathered from 18 subjects post-stroke and 8 control subjects (Cruz et al.) was used to produce curves of pelvic obliquity for both groups, as shown in Figure 68. The mean walking speed of the patients was 1.8 km/h. Therefore, the protocol was designed to be performed at the same walking speed. Interaction forces between the exoskeleton worn by the subjects and the RGR Trainer s actuation system during left leg swing were used as the outcome measure. 87

102 Figure 68: Pelvic obliquity trajectories collected from healthy and impaired subjects in the study by Cruz et al. [44]. The impaired subjects (stroke) clearly exhibit a hip-hiking pelvic motion trajectory Protocol 1 Details 1. Subject walked at 1.8 km/h on the treadmill inside the RGR Trainer and selected his comfortable cadence at this speed. Walking speed was dictated by that found by Cruz et al. study. The actuation system operated under zero force control (back-drivable mode), minimizing interaction forces and allowing for maximum freedom of movement. 2. Baseline pelvic obliquity and hip and knee joint angles were collected over 100 strides and converted into the baseline pelvic obliquity reference trajectory and synchronization reference respectively, by segmenting the data according to heel strikes (as detected by a foot switch in the subject s left shoe) and averaging across all gait cycles. 88

103 The following 4 time epochs played out to form a continuous run, as outlined in Figure 69. Protocol 1 Baseline Hip-Hike Baseline Backdrive 1 minute Force Field On 1 minute Force Field On 3 minutes Force Field On (Error Clamp) 1 minute Backdrive 1 minute Epoch 1 Epoch 2 Epoch 3 Epoch 4 Epoch 5 Figure 69: Graphical representation of protocol 1. Throughout the test, force, position and gait cycle location data were recorded continuously. 3. Epoch 1. Duration: 1.0 minutes. The subject walked freely (back-drivable mode) on the treadmill at the previously specified speed, with a metronome setting the cadence. The robot synchronized to the subject s gait by using the subjects own reference synchronization trajectory (8-DOF). 4. Epoch 2. Duration: 1.0 minutes. The force field was activated, with the subject s own baseline still serving as reference trajectory. 5. Epoch 3. Duration: 3.0 minutes. Adaptation Period. The reference trajectory was switched from baseline to the hip-hiking pattern. 6. Epoch 4. Duration: 1 minute. De-adaptation Period. The position reference was switched to subject s own baseline, with the force field still active. 89

104 7. Epoch 5. Duration: 1 minute. The force field was switched off (backdrivable mode) Protocol 1 Results Various levels of force field strengths (K c ) were tested on four subjects, ranging between 6 and 24N-m/deg., as this was an exploratory series of experiments. The force field of 12 N-m/deg produced the most distinct response from one particular subject (Figure 70). As the reference trajectory was switched from baseline to hip-hiking with the force field active, the magnitude of interaction forces spiked up to 200N (relative to the average force level in the first epoch of the test), and decreased as the subject adopted a new, altered gait pattern. The exponential decay trend line fitted to the interaction force data shows that the subject altered his gait pattern, reducing these interaction forces, with a time constant of 152 gait cycles (+/- 130). Following the 3-minute training session, the position reference was switched back to baseline, which resulted in another spike in the interaction forces. Figure 70 displays force magnitude (absolute value), but in fact the interaction force switched sign at that instant, pushing the subject s hip down. In other words, the subject continued to hip-hike for several gait cycles, as the device attempted to retrain the subject to walk with his own baseline gait pattern. The short 1 minute period was not quite long enough to reach a steady-state level. 90

105 Figure 70: Result of hip-hike inducing test. The interaction force magnitudes measured by the load cell (F int ) are graphed. Models fitted to the data (Fit) were used to estimate time constants Protocol 1 Discussion The test results were very encouraging, indicating that a short training session in the RGR Trainer could lead to learning of a new gait pattern through alteration of the pelvic obliquity angle while walking, and that this newly learned gait pattern can persist for a short period of time. Regarding force field strength, we observed that at K c =6N-m/deg the subjects had difficulty perceiving the gait pattern which the RGR Trainer was attempting to impose on them. At K c =24N-m/deg one subject reported exploring the haptic feedback by varying his pelvic obliquity across several gait cycles, and being unable to perceive variations in 91

106 the force, possibly due to saturation. Reducing the force field strength to K c = 12Nm/deg seems to have remedied this problem, and this force field strength produced the result shown in Figure 70. Finding the optimal force field strength to maximize retention became one of the goals in subsequent experiments Protocol Hypothesis The success of gait retraining via force fields applied to pelvic obliquity can be judged by the retention of the kinematics of the newly learned gait pattern in unrestricted gait following the training period. This is in contrast to the method used in Protocol 1, where the retention of the newly learned gait pattern was ascertained via the interaction forces during the un-learning epoch Protocol 2 Methods Testing Approach and Outcome Measure As an outcome measure, the retention of the newly learned gait pattern kinematics following the training epoch was used. After the training epoch (hip-hiking position reference and active force field), the system was switched to backdrivable mode, with an expectation that the subject would continue to exhibit a hip-hiking gait pattern for some time. In an attempt to make the switch between modes less perceivable by the subject, a tunnel around the hip-hiking reference trajectory was used. A similar tunnel has been used by other teams, for example in [45] and [40]. The tunnel was physically implemented in the controller by nullifying the position error while it was less than a 92

107 particular value (tunnel semi-width), and once the position error surpassed the tunnel semi-width, it was offset by that value Regarding walking speed used, a split-belt treadmill adaptation study [46] suggests that there is a partial separation in the functional networks controlling fast and slow walking, and the rate of transfer of treadmill gait retraining effects to over-ground walking is therefore expected to be higher if it is performed at the same walking speed. Hence, in Protocol 2, the subjects selected their own comfortable walking speed (CWS), at which the experiments were performed. 1. Subject walked at his CWS on the treadmill inside the RGR Trainer and selected his own cadence at this speed. The actuation system operated under zero force control (backdrivable mode), minimizing interaction forces and allowing for maximum freedom of movement. 2. Baseline pelvic obliquity data and hip and knee joint data were collected and converted into the baseline pelvic obliquity reference trajectory and synchronization reference respectively by segmenting the data according to heel strikes (as detected by a foot switch in the subject s left shoe) and averaging across all gait cycles. 3. Subject walked again at his CWS, while performing a simulated hip-hiking gait pattern. A mean representative hip-hiking trajectory for that particular subject was extracted and used in the subsequent test. 93

108 Protocol 2 The following 4 test epochs played out to form a continuous run, as outlined in Figure 71. The epoch durations were based on the number of gait cycles completed, as opposed to the time elapsed as was done in Protocol 1. Throughout the test, the interaction force, pelvic obliquity angle and gait cycle location data were recorded continuously. 4. Epoch 1. Duration: 100 gait cycles. The subject was allowed to walk freely on the treadmill at their previously found CWS. 5. Epoch 2. Duration: 100 gait cycles. With the subject s baseline pelvic obliquity as the position reference, and with the tunnel size set at 1 degree (half-span), the force field was activated. 6. Epoch 3. Duration: 300 gait cycles. The reference trajectory was switched from the subject s own baseline to the hip-hiking trajectory. 7. Epoch 4. Duration: 100 gait cycles. The force field was switched off. This epoch differs from that used in Protocol 1, since the subject was not forced to switch back to own baseline (error clamp), but was given freedom to continue walking with the newly-acquired gait pattern. This epoch was used to record the outcome of gait retraining, which occurred in epoch 3. 94

109 Protocol 2 Baseline Hip-Hike Force Field On Force Field On Backdrive 100 cycles 100 cycles (train) 300 cycles Backdrive 100 cycles Epoch 1 Epoch 2 Epoch 3 Epoch 4 Figure 71: Graphical representation of protocol Protocol 2 Results In Protocol 2, the outcome measure was the degree of hip-hike in the subject s pelvic obliquity immediately following the hip-hike training epoch. In the top graph of Figure 72, the solid red curve (ref HH ) is the hip-hiking gait pattern from Cruz et al. The action of the force field during the training period (epoch 2) resulted in this particular subject exhibiting a hip-hike, the mean of which across all gait cycles in epoch 2 is marked in dash-dot red curve (act HH ). In the bottom graph we see all of the subject s obliquity curves from all the gait cycles in epoch 4, displayed in dashed blue. 95

110 Obliquity [degrees] Obliquity [degrees] Protocol 2 Result ref BL ref HH act BL act HH act AE Percent Gait Cycle Percent Gait Cycle Figure 72: Sample result from one subject tested under Protocol 2. In the top graph, ref BL is the baseline pelvic obliquity of the subject, ref HH is the hip-hiking trajectory from Cruz et.al, act BL is the mean pelvic obliquity trajectory under force field, act HH is the average hip-hiking trajectory produced by the subject under force field and act AE is the average pelvic obliquity following hip-hike training session. In the bottom graph, pelvic obliquity curves from all gait cycles in epoch 4 are shown (dashed blue) along with baseline (solid black). Here one gait cycle spans between consecutive left foot toe-offs Protocol 2 Discussion The goal of the above-described Protocol 2 experiment was to assess the RGR Trainer s ability to teach the hip-hiking gait pattern by the retention of the newly-learned kinematics. By employing a tunnel around the hip hiking reference which came from 96

111 the subject itself, it was expected that the perceptibility of force field removal would be reduced, as long as the subject tracked the hip-hiking pattern closely during the training epoch. Running this protocol at faster walking speeds meant that the hip-hiking obliquity reference from Cruz et.al. could not be used anymore. Unfortunately, since the hip-hiking obliquity references were collected from the subjects themselves, they were not naïve to the goals of the experiment anymore. Also, despite the use of a +/- 1 degree tunnel around the hip-hiking reference, the subjects recognized immediately the removal of the force-field. We attempted to address these issues in Protocol Protocol Protocol 3 Preliminary Tasks In order to enable testing of naïve subjects at their own CWS (ca. 3km/h) with a hiphiking reference trajectory which would match their own simulated hip-hiking pattern better than that found in Cruz et al. (collected at only 1.8km/h), it became necessary to generate a representative hip-hiking trajectory (simulated by healthy subjects) collected at the higher walking speed Representative Simulated Hip-hiking Obliquity Trajectory Left and right hip-hike pelvic obliquity trajectories were collected using the RGR Trainer under two conditions: with visual feedback and without. The tests with visual feedback ensured that the subjects performed a hip-hike of predetermined magnitude, while the tests without visual feedback ensured that a more representative variability in maximum simulated hip-hike angle in healthy subjects was found. Sample plots of baseline pelvic 97

112 obliquity, and hip-hiking obliquity under two conditions (with visual feedback and without) from five of the subjects are shown in Figure 73 through Figure 77. These five subjects all had their right foot as the dominant foot. Figure 73: Subject # 2 baseline and hip-hiking plots. Baseline and its standard deviation curves are plotted along with their inverses (with first half of gait cycle plotted first, and vice versa) in order to aid in visualizing symmetry. Right hip-hike curves are also inverted and plotted in reverse order to facilitate comparison with left hip-hike curves. 98

113 Figure 74: Subject # 3 baseline and hip-hiking plots. Figure 75: Subject # 4 baseline and hip-hiking plots. 99

114 Figure 76: Subject # 7 baseline and hip-hiking plots. Figure 77: Subject # 8 baseline and hip-hiking plots. 100

115 Visual feedback helped the subjects produce very similar maximum hip-hike angle magnitudes, even when there was an offset present in the obliquity angle as in the cases of subjects 4 and 7 (Figure 73 and 77 ). This suggests that the subjects relied heavily on visual feedback presented to them. On the other hand, when visual feedback was not used, the subjects tended to overshoot the maximum hip-hike. Unequal inertia on the left versus right side of the actuation system may be the cause of this result. Subjects hiphiked with the left side first, and when performing the same action with the right side, it is possible that under no-visual-feedback conditions they applied an internal model of motor control of their left hip side onto the opposite side (right). All but one subject (#7) in Table 1 were right-foot-dominant. Table 1: Summary of hip-hike data. The means exclude subject # 2 data. Only one subject (#6) was left-foot-dominant. Hip Hike Base - Left w/ Left w/o Right w/ Right w/o line Feedback Feedback Feedback Feedback Subject St. St. St. St. St. Angle Angle Angle Angle Dev. Dev. Dev. Dev. Dev Mean The standard deviation figures coming from the hip-hiking gait patterns are the maximum values found in the portions of the gait cycle when the hip-hike occurs, i.e. ~ 55% to 90% 101

116 for left, and 10% to 45% for right hip-hike. The right hip-hike curves were inverted and their order was reversed (first half of gait cycle versus second half) in order to make it easier to compare them to the left hip-hike curves. Interestingly, the standard deviation values for left and right maximum hip-hike angle were very similar in both cases: with visual feedback and without. In other words, precision was quite similar, while accuracy quite different between the two conditions. Figure 78: Mean pelvic left hip-hike obliquities of 8 subjects (across ca. 200 gait cycles) and mean across the means of 7 subjects (subject # 2 data excluded due to the timeseries extreme mismatch with the rest due to erroneous foot switch operation). The wide range of peak hip-hike angle distribution among the gait cycles shown in Figure 78 above stems from the fact that subjects walked at self-selected CWS and cadence. 102

117 Visual feedback of maximum hip-hike angle would not be available to the subjects during hip-hike training tests, therefore the mean of single standard deviations found in the tests without visual feedback was used to select the tunnel size, while employing the hip-hiking obliquity from Figure 78, which was based on tests with visual feedback, hence featuring the appropriate maximum hip-hike magnitude of about 6 degrees. The mentioned mean of standard deviations of maximum hip-hike angle across 7 subjects was found to be 1.61 degrees. Therefore, it was speculated that the majority (approx. 68%) of all the left hip-hiking gait cycles (without visual feedback) exhibited a maximum hip-hike angle which fell within +/ degrees of the mean. Thus we set the tunnel width to +/- 1.6 degrees Protocol 3 Methods Outcome Measure Selection of optimal force field strength was done based on variability of maximum hiphike angle. The optimal force field strength would be that of lowest magnitude, which would produce the hip-hike angle variability σ 2 2 which was found to be statistically indeterminate from that of the reference trajectory. F-test null hypothesis: H o : σ 1 2 = σ 2 2 (5.1) and the alternative hypothesis: H 1 : σ 1 2 σ 2 (5.2) with a level of significance: 103

118 α = 0.05 (5.3) we get the following f values: f 0.05 (199,199) 1.25 (5.4) and: f (199,199) 0.8 f (199,199) 1.25 (5.5) Therefore, the null hypothesis is rejected when either f > 1.25 or if f < 0.8, where: 0.05 f s (5.6) s To confirm our assumptions, the power of the test was analyzed with G*Power software ( An F-test of equality with a detectable ratio of variances of 1.4, error probability α = 0.05, allocation ratio N1/N2 = 1 and both sample sizes of 197 each produces the power of the test of The ratio of variances of 1.4 is equivalent to a ratio of standard deviations of Protocol 3 Setup 1. Subject walked at his CWS on the treadmill inside the RGR Trainer and selected his own cadence at this speed. The actuation system operated under zero force control (backdrivable mode), minimizing interaction forces and allowing for maximum freedom of movement. 2. Baseline pelvic obliquity data and hip and knee joint data were collected by recording the RGR Trainer s pelvic brace position measurement and hip and knee angle measurements, and converted into the baseline pelvic obliquity reference 104

119 trajectory and synchronization reference respectively by segmenting the data according to heel strikes (as detected by a foot switch in the subject s left shoe), and averaging across all gait cycles Protocol 3 Details Five different force field levels (K c = 20, 25, 30, 35 and 40N-m/deg) were randomized. 3. Epoch 1. Duration 50 gait cycles. The subject ambulated, and reached a steady state pace. 4. Epoch 2. Duration 300 gait cycles. Subject was exposed to a force field selected randomly out of 5, with the representative hip-hike pattern serving as the position trajectory, activated between 55% and 85% of the gait cycle. A previous test result (Figure 70) suggested a time constant of about 150 gait cycles, therefore here twice the time-constant was used: 300 gait cycles. 5. Epoch 3. Duration 200 gait cycles. The 200 cycles were used to record the outcome of gait retraining. 6. Epoch 4. Duration 300 gait cycles. An error-clamp setting, with tunnel set to +/- 0.7 degrees and a force field of K = 30 N-m/deg was used. 0.7 degrees is the single standard deviation of the mean of baseline obliquities of seven subjects from Table Epoch 5. Duration 200 gait cycles. Force field was off. Obliquity from this epoch was used to confirm that de-adaptation was sufficient. One potential method was computing the sample variance s 2 2 and performing an F-test against reference (baseline) variance. 105

120 Steps 2 through 5 were repeated for the other four levels of force field strength. The total duration of the test was 50 + (1000 x 5) = 5050 strides. The timing of force field activation (sigmoid switch) was carefully picked to occur after the initial reversal of the pelvis direction of motion had occurred (from pelvic drop to hip-hike). Testing revealed that the interaction force during that initial direction reversal was found to be highly indicative of the device s operating mode. Protocol 3 Hip-Hike Baseline Force Field On Force Field On Backdrive 50 cycles (train) 300 cycles Backdrive 200 cycles (error clamp) 300 cycles Backdrive 200 cycles Epoch 1 Epoch 2 Epoch 3 Epoch 4 Epoch 5 Figure 79: Graphical representation of protocol 3 trial. A session consisted of five such trials concatenated into a single run. Epoch 1 was only used in trial 1 of the run, and each trial used a different force field magnitude Protocol 3 Results The three subjects tested under the protocol presented here were all naïve to the operation of the device. Great care had been taken in selecting the size of the tunnel around the hip-hiking reference trajectory such that the subjects would not be able to sense the transition between the hip-hike train epoch (2) and the subsequent backdrive epoch (3). The line of reasoning was that in the majority of gait cycles during the experiments, the subjects would produce a hip-hike angle which would fall within the tunnel, and hence upon entering epoch 3, when the force field was removed, the subjects would not 106

121 Obliquity [degrees] be able to recognize a change in the operating mode of the RGR Trainer. Yet, despite these measures, the subjects were able to tell immediately when the operating mode of the device changed. Sample results from three subjects at selected force-field strengths are presented below, while the complete results for all three subjects can be found in Appendix B. Subject 1-30N-m/deg ref BL -2 ref HH act FF -4 act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 80: Taking into account the offset in pelvic obliquity between baseline (ref BL ) and pelvic obliquities resulting from hip-hike training, an after-effect can be observed (act AE- Early) which diminishes with time (act AE-Late ). 107

122 Obliquity [degrees] Subject 2-40N-m/deg ref BL ref HH act FF act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 81: Subject 2 exhibits significant after-effect, characterized by exaggerated pelvic drop at the beginning of the third epoch (backdrive), labeled as act AE-Early. This aftereffect diminishes throughout the duration of the third epoch, and the de-adapt epoch (4 th ) seems to accomplish its task (act de-adapt ). 108

123 Obliquity [degrees] Subject 3-30N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 82: This subject s baseline pelvic obliquity (ref BL ) has a gross offset. Nevertheless, the action of the RGR Trainer does make the subject produce a hip-hike during the second half of the gait cycle. Unlike what was observed in Figure 80 and 81, the subject exhibits a reduced pelvic drop, which returns to baseline over time Protocol 3 Discussion The RGR Trainer was able to coax all three subjects to exhibit a hip-hiking gait pattern when the force field was activated, judging by the mean pelvic obliquity during epoch 2 (act FF ). In epochs 2 and 3 the subjects did not respond to this protocol as predicted. Instead of oscillating within the tunnel around the hip-hiking trajectory in epoch 2, the subjects tended to stay at the bottom of the tunnel: peaks of actual hip-hiking profiles under the action of force field (act FF ) are approximately 2 degrees lower than hip-hiking reference ref HH, which is close to the semi-width of the symmetrical tunnel around the reference trajectory (1.6 degrees). Then, in epoch 3, as the force field was removed, the 109

124 subjects gradually stopped hip-hiking, but this change did not occur subconsciously, because the subjects immediately realized that the force field was removed. Therefore, it was biased by the subjects motivation and understanding of their task. As a result of the subjects hip-hiking at the bottom of the reference tunnel and interacting with the force field, the removal of the force field in epoch 3 produced a sensation of a downwarddirected force, which in turn caused the subjects to exhibit an increased pelvic drop, as compared to their baseline pelvic obliquity. In other words, while these experiments were assistive in nature, and the subjects were instructed to follow the guidance, the subjects adapted to the forces present in epoch 2, and they de-adapted in epoch 3, i.e. they underwent motor adaptation Protocol Introduction Protocols 2 and 3 attempted to quantify the effectiveness of assistive gait retraining strategies with the retention levels of the newly acquired gait pattern. Implementing a protocol which measures gait pattern retention which is unbiased by subjects motivation proved to be very challenging. Protocol 4 aimed to investigate motor control of the pelvis by comparing both assistive and resistive training Protocol 4 Methods Four healthy subjects participated in the study. Each subject completed two assistive and two resistive training sessions. During the assistive training sessions, the subjects were instructed to follow the guidance of the RGR Trainer, and during the resistive sessions, the subjects were instructed to maintain their own natural gait pattern and not to allow the 110

125 RGR Trainer to alter it. For each training type, two variations of epoch 3 were used: backdrive and playback. In the backdrive epoch (epoch 3b), the actuation system operated in backdrivable mode, while in the playback epoch (epoch 3p), the mean commanded force profile from the last ten gait cycles of epoch 2 (the epoch immediately preceding epoch 3p) was played back throughout the duration of epoch 3p. Therefore, while the subjects were exposed to a force field in epoch 2, in epoch 3p they were exposed to a constant force profile, which was only a function of the subject s temporal progression through the gait cycle, and not a function of their pelvic obliquity angle. Each session consisted of three trials, with each trial testing one of three force field magnitudes (5, 15 and 25N-m/deg), randomized in order. Each trial lasted 1200 strides, and consisted of four 300-stride epochs: hip-hike train (epoch 2), backdrive (epoch 3b) or playback (epoch 3p), error clamp (epoch 4), and backdrive (epoch 5). Whether a trial used a backdrive mode or playback mode was randomized, but ensuring that each combination of operating mode and force field magnitude was covered. Each session began with a 50-stride initiation epoch, hence each session lasted 50 + (3*1200) = 3650 strides. In protocol 4 a tunnel around the hip-hiking reference was not used. The force field activation switch was set to go on at 44% of the gait cycle (coinciding very closely with toe-off) and to go off at 76% (in order to diminish to zero by left heel strike the end of the gait cycle see Figure 60) Protocol 4 Setup 1. Subject walked at his CWS on the treadmill inside the RGR Trainer and selected his own cadence at this speed. The actuation system operated under zero force control (backdrivable mode). A metronome was set to the subject s cadence. 111

126 2. As the subject ambulated for 100 gait cycles to the cadence set by the metronome, baseline pelvic obliquity timeseries and hip and knee joint angle timeseries were collected and converted into the baseline pelvic obliquity reference trajectory and synchronization reference respectively, by segmenting the data according to heel strikes (as detected by a foot switch in the subject s left shoe), and averaging across all gait cycles Protocol 4 Details 3. With the metronome setting the cadence, the subject ambulated in the RGR Trainer for 5 minutes, with the system in backdrivable mode in order to reach steady state. 4. Epoch 1. Duration 50 gait cycles. Initiation: every 5 strides, the RGR Trainer switched between two operating modes: error clamp (baseline reference) and hiphike train. This was done to make the subjects accustomed to the operation of the system, and to make subjects believe that there are only two operating modes. 5. Epoch 2. Duration 300 gait cycles. Subject was exposed to a force field selected randomly out of 3, with the representative hip-hike pattern serving as the position trajectory, activated between 44% and 76% of the gait cycle. 6. Epoch 3b. Duration 300 gait cycles. The system operated in backdrivable mode. 7. Epoch 3p. Duration 300 gait cycles. The system operated in playback mode, generating a constant force profile (mean commanded force from last 10 gait cycles in epoch 2) as a function of temporal progression through the gait cycle. 112

127 8. Epoch 4. Duration 300 gait cycles. Error-clamp (K c =15N-m/deg) with subject s own baseline trajectory was used to de-adapt the subject. 9. Epoch 5. Duration 200 gait cycles. Backdrive mode. Pelvic obliquity during this epoch could be used to confirm de-adaptation. Protocol 4 Hip-Hike Hip-Hike Baseline Force Field On Force Profile Playback Force Field On (train) 300 cycles Backdrive 300 cycles (error clamp) 300 cycles Backdrive 300 cycles Epoch 1 Epoch 2 Epoch 3b/3p Epoch 4 Epoch 5 Figure 83: Graphical representation of a single trial of protocol 4. A complete session consisted of three trials running continuously (each using a different force field), with trial 2 and 3 consisting of epochs 2 through 5. Epoch 3 type was randomized. Two sessions per training type (assistive or resistive) were run to ensure that every combination of force field and epoch 3 type was tested Protocol 4 Results Assistive Training: Hip-Hike to Backdrive Results of assistive training followed by backdriven mode across all four subjects and for the three force field strength levels are shown in Figure 84 through 89. All subjects hiphiked during the training epoch, and a marked increase in hip-hike magnitude can be seen between 5 and 15N-m/deg force field strengths. 113

128 Figure 84: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at K c = 5N-m/deg) followed by epoch 3b (backdrive, check-). Subjects hip-hiked during the training period, and exaggerated pelvic drop in the subsequent epoch may indicate motor adaptation. 114

129 Figure 85: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 15N-m/deg) followed by epoch 3b (backdrive, check-). Subjects hip-hiked during the training period. The exaggerated pelvic drop in the subsequent epoch (3b) may be a sign of motor adaptation. 115

130 Figure 86: Assistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 25N-m/deg) followed by epoch 3b (backdrive, check-). Subject 2 exhibited most exaggerated pelvic drop in epoch 3b. The cosine distance plots (Figure 87 through 89) compare the pelvic obliquity during the hip-hike training epoch and the backdrive epoch to their respective reference trajectories, i.e. for strides the pelvic obliquity is compared to the hip-hiking reference, and during strides the pelvic obliquity is compared to the baseline reference. These figures show that upon switching from the training epoch to the backdrive epoch, the subjects stopped hip-hiking immediately, and began exhibiting a gait pattern very similar to their baseline. 116

131 Figure 87: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 5N-m/deg) and to baseline reference ( strides, epoch 3b) during assistive training. 117

132 Figure 88: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 15N-m/deg) and to baseline reference ( strides, epoch 3b) during assistive training. Only subject 4 exhibited a gait pattern close to hip-hiking for several gait cycles after the switch occurred. 118

133 Figure 89: Similarity of pelvic obliquity to hip-hiking reference (0-300 strides, training at 25N-m/deg) and to baseline reference ( strides, epoch 3b) during assistive training Resistive Training: Hip-Hike to Backdrive Results of resistive training followed by backdriven mode across all four subjects and for the three force field strength levels are shown below. Figure 90 through 92 display pelvic obliquity data during epoch 2 and epoch 3b. In general the four subjects had difficulty resisting the device, and so their pelvic obliquity during epoch 2 (hip hike training epoch) rather closely resembles the hip-hike reference. Subject 1 was able to resist the guidance 119

134 of the device much better than the other subjects, but only at the two weaker force field magnitudes. Figure 90: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 5N-m/deg) followed by epoch 3b (backdrive-check). 120

135 Figure 91: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 15N-m/deg) followed by epoch 3b (backdrive-check). 121

136 Figure 92: Resistive training: baseline trajectories and pelvic obliquities during epoch 2 (train at 25N-m/deg) followed by epoch 3b (backdrive-check). 122

137 Figure 93 through 95 show the mean interaction force profiles during epochs 2, 3 and 4. Subject 1 resisted the force fields of 5 and 15N-m/deg much more than the other 3 subjects, and this explains his lowest pelvic obliquity angle among all the subjects during hip-hike training, as shown in Figure 90 and 91. That same subject reported not recalling what his baseline obliquity pattern was during the test. Interestingly, in the error clamp epoch this subject resisted the device by continuing to exhibit a low amplitude hip-hike. Figure 93: Resistive training at 5N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp). 123

138 Figure 94: Resistive training at 15N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp). Subject 1 resisted the force field the most, peaking at almost 100N (mean), and during the Error Clamp epoch he again fought against the system the most (-40N mean). This particular subject reported verbally during the training session having difficulty recalling his baseline pelvic obliquity pattern. 124

139 Figure 95: Resistive training at 25N-m/deg: Interaction forces during epoch 2 (Hip Hike), epoch 3b (Backdrive), and epoch 4 (Error Clamp) Hip-hike to Playback In the trials in which the training epoch was followed by the force playback epoch (3p), the subjects were unaware of any transition in operating mode taking place. Since a force field was replaced by a constant force profile, the variability in hip-hike angle did increase in the playback epoch. 125

140 Protocol 4 Discussion Assistive Training followed by Backdrive Though the combination of assistive hip-hike training followed by backdrivable epoch had been used in Protocol 3, it was repeated here so that the assistive and resistive training paradigms could be compared and contrasted. This time no attempt was made to conceal the transition between the two epochs; no tunnel was used, and the sigmoid was turned on at 44%, coinciding with the left toe-off, which resulted in increasing the effect that force field had on hip-hike generation. The RGR Trainer caused all subjects to hiphike during epoch 2 (Figure 84 through 86), but turning the force-field off resulted in immediate switch to baseline obliquity gait pattern (Figure 87 through 89). In fact in the majority of the trials the subjects exhibited an exaggerated pelvic drop, and all subjects stated that the device was pushing them down when the system switched from the training epoch to the backdrive epoch. This sensation may be explained by the subjects both getting accustomed to the haptic feedback and by the subjects adopting a new gait pattern during the training period. This result suggests that some subjects experienced motor adaptation and de-adaptation. Motor adaptations in studies are usually evoked through resistive training [37, 38, 40, 42, 43], but the fact that during assistive training the subjects still experienced significant forces (judging by position error in Figure 84 thru 86) could explain why there are signs of motor adaptations. The significant position error in assistive training also suggests that hip-hiking up to 6 degrees of rotation makes the gait pattern very inefficient and uncomfortable. 126

141 Resistive Training followed by Backdrive Resistive training produced rather surprising results. Despite being instructed to not allow the device to alter their gait pattern, in the majority of the trials the subjects did hiphike. One notable exception is the trial at 5N-m/deg with subject 1. This subject was able to keep the hip-hike angle to a much lower magnitude than the other subjects. Though in assistive training we observed a trend of increased pelvic drop in the backdrive epoch among some subjects and trials, resistive training seems to have had the opposite effect, although it is not consistent across all subjects and all force field strength levels. There seems to be a correlation between the effort put into resisting hip-hiking during epoch 2 and the magnitude of hip-hike angle in epoch 3b. Subjects 1 and 3 produced the smallest hip-hike magnitudes during epoch 2, and subsequently produced the smallest pelvic drop magnitudes in epoch 3b (i.e. they exhibited a small hip-hike) Conclusion The four test protocols presented in this chapter describe the evolution of approaches, which were used to investigate the RGR Trainer s ability to administer gait retraining to healthy subjects. Experiments with all four protocols demonstrated the system s ability to alter healthy subjects gait by applying a moment to pelvic obliquity to cause hiphiking, and signs of retention of the gait patterns being trained were observed in many of the subjects tested with the four protocols. Retention of newly learned motor patterns has been stated as the proof of motor learning in experiments with upper and lower extremities [37, 38, 40-43], and studies with stroke survivors [37, 43] have found that adapting the patient to a perturbation that worsened or amplified their error was what 127

142 drove adaptation to result in after-effects that improved their movement. The experiments with protocols 1 through 3 and half of the experiments with protocol 4 were assistive in nature, where the subjects were instructed to let the device guide their pelvis. Only protocol 4 included resistive training. It is the resistive training paradigm which most motor learning studies have relied on to assess motor adaptations. In resistive training of protocol 4, subjects were instructed to resist the guidance of the RGR Trainer (which was imposing on them the hip-hiking gait pattern) and to maintain their own gait pattern, which means exhibiting pelvic drop during leg swing. The results of resistive training presented here indicate that in the majority of the trials the subjects were not actively engaging in exhibiting pelvic drop in their gait pattern. It may be the case that pelvic drop occurs passively, due to upper body s weight, which would mean that humans don t normally engage muscles of the lower body to drive the pelvic drop. Therefore, the results of the study using protocol 4 in regards to motor control of the pelvis are not conclusive, and further testing with redesigned protocols will have to be performed in the future. 128

143 Conclusions 1.38 Summary Stroke is a leading cause of disability in America, severely affecting victim s locomotion. Manual gait rehabilitation is costly and physically demanding, which led to an increased interest in robotic devices. Nevertheless, one limitation which the commercially available robotic devices for rehabilitation seem to share is constraint of the natural pelvic motions. Addressing of both primary gait deviations in the legs as well as the secondary gait deviations in the pelvis (which arise as compensatory movements to the primary gait deviations) may offer a much needed improvement in the outcome of gait retraining following stroke. Therefore, the Robotic Gait Rehabilitation (RGR) Trainer was designed and built to be a simple, low cost device with a single actuator, which allows all the natural motions of the pelvis, while being able to selectively and compliantly guide the pelvis in the frontal plane (pelvic obliquity) in order to target hip-hiking in patients post stroke. This device uses impedance control and human-machine synchronization to generate corrective forces as a response to deviations from pre-determined pelvic obliquity trajectories. The corrective forces are transferred to the subject via a lower body exoskeleton, which can very effectively transfer forces to the pelvis, while its 10 DOFs allow for unhindered ambulation on the treadmill. Testing of several healthy subjects using four different protocols has demonstrated that the RGR Trainer can effectively alter the gait by imposing a hip-hiking gait pattern, and 129

144 retention of this gait patter has been observed following the training sessions, which implies learning. The qualities of the RGR Trainer make it a very good tool for the study of motor control of the pelvis, the understanding of which is expected to lead to designing better gait retraining strategies in the future Future Work Mechanical Design One minor hardware upgrade, which may improve both the range of motion of the actuation system and the performance of the control system is the replacement of the linear potentiometer, which is used with the linear actuator position measurement to compute pelvic obliquity angle. Installation of a digital linear encoder with proper range of motion would eliminate the electrical noise inherent to analog devices, and therefore possibly increase the operating range of the impedance controller. The first major mechanical upgrade of the system will be the addition of an assembly, which is referred to as the horizontal motion system in section This assembly increases the range of lateral motion of the subject in the device, and it shifts the location of the subject forward of the main frame. This will give a physical therapists unrestricted access to the patients legs for assisting with stepping and knee flexion. Extending the operating envelope of the RGR Trainer to pelvic rotation will require another major upgrade of the mechanical system. One possible design, which uses four linear actuators to control 2 rotational DOFs of the pelvis is presented in Appendix C - RGR Trainer 2DOF. 130

145 Control Strategies and Software Many unforeseeable problems are likely to arise due to testing of patients post-stroke in the RGR Trainer, but one issue which is very likely to appear will be lower accuracy of the human-machine synchronization system s gait cycle location estimate. This problem is expected to arise mainly due to pronounced gait asymmetries in patients, as well as a more dramatic modification of gait parameters within a training session. Therefore, the human-machine synchronization algorithm will have to be developed further, for example by the addition of weight assignment to the 8 DOFs currently used, such that the system becomes more resilient to perturbations Human Testing Healthy subject testing presented in 0 demonstrated the RGR Trainer s ability to effectively and reliably apply corrective force fields to pelvic obliquity. This work has laid down a foundation for future experiments, which will try to answer more directly questions about motor control of the pelvis. These experiments will in turn allow for further testing to take place with patients post-stroke, which will demonstrate the efficacy of addressing secondary gait deviations in gait retraining. 131

146 Appendix A - Impedance Controller Bench Test Results Figure 96 to 111 presented here document the process of bench-testing the impedance controller implemented for linear motion of the servo-tube actuator, as described in section Figure 96: Position error (Des. Position Act. Position) generated the force command (F virt ). Load cell measured actual interaction force (F ext ). 132

147 Figure 97: The addition of damping attenuated the oscillatory force interaction. Effects of stiction can be seen just past maxima and minima. Figure 98: Higher gain value caused greater environment deflection (Act. Position). Lack of damping again resulted in oscillatory response. 133

148 Figure 99: Once again, after the damping ratio (zeta) was introduced, the oscillatory behavior diminished. Figure 100: At this high virtual stiffness setting, slight vibrations were again felt, and can be seen in the F ext signal. 134

149 Figure 101: Again, with the damping ratio increased, the vibrations diminish. Figure 102: As the damping ratio was increased to 0.6, undesirable behavior appeared. The virtual damper component of the command signal began displaying vibratory behavior. The resulting forces were felt by the subject, but are not present in the measured force signal F ext due to low-pass filtering. 135

150 Figure 103: With the reference trajectory of 3Hz and no damping, the measured force signal (F ext ) tended to lag behind the commanded force (F virt ). Figure 104: Increase in damping ratio smoothed out both force curves. 136

151 Figure 105: With the damping ratio set to 0.5, the system still behaved well. Figure 106: Once the damping ratio was increased to 0.6, the performance deteriorated due to appearance of high frequency vibrations, which can be seen in the F virt signal, and could be felt by the subject. 137

152 Figure 107: With the reference trajectory frequency increased to 6Hz, actuator thrust rod s inertia caused significant distortions to the position and force profiles. The system still behaved in a stable manner. Figure 108: Introduction of damping had the effect of correcting the profile of the external measured force F ext, by properly modulating the virtual force F virt. 138

153 Figure 109: Inertial effects cause the measured force profile F ext to lag significantly behind position error. Figure 110: Increasing the damping ratio seemed to make the controller efforts (F virt ) more abrupt. 139

154 Figure 111: As seen before, damping ratio of 0.6 amplified the derivative action of the PD impedance controller. This caused high frequency vibrations. 140

155 Obliquity [degrees] Appendix B - Protocol 3 Results Trials (5 per subject) are listed for the three subjects, ordered by the force field strength used (ascending), and with the sequence number of the particular trial listed in figure caption (Figure 112 through 126). Subject 1-20N-m/deg ref BL -2 ref HH act FF -4 act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 112: Subject 1 2 nd trial. 141

156 Obliquity [degrees] Obliquity [degrees] Subject 1-25N-m/deg ref BL ref HH act FF act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 113: Subject 1 3rd trial. Subject 1-30N-m/deg ref BL -2 ref HH act FF -4 act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 114: Subject 1 4th trial. 142

157 Obliquity [degrees] Obliquity [degrees] Subject 1-35N-m/deg ref BL ref HH -4 act FF act AE-Early -6 act AE-Late act de-adapt Percent Gait Cycle Figure 115: Subject 1 1st trial. Subject 1-40N-m/deg ref BL ref HH act FF -4 act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 116: Subject 1 5 th trial. 143

158 Obliquity [degrees] Obliquity [degrees] Subject 2-20N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 117: Subject 2 5 th trial. Subject 2-25N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 118: Subject 2 2 nd trial. 144

159 Obliquity [degrees] Obliquity [degrees] Subject 2-30N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 119: Subject 2 1 st trial. Subject 2-35N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 120: Subject 2 3 rd trial. 145

160 Obliquity [degrees] Obliquity [degrees] Subject 2-40N-m/deg ref BL ref HH act FF act AE-Early act AE-Late -6 act de-adapt Percent Gait Cycle Figure 121: Subject 2 4 th trial. Subject 3-25N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 122: Subject 3 1 st trial. 146

161 Obliquity [degrees] Obliquity [degrees] Subject 3-30N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 123: Subject 3 2 nd trial. Subject 3-35N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 124: Subject 3 3 rd trial. 147

162 Obliquity [degrees] Obliquity [degrees] Subject 3-40N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 125: Subject 3 4 th trial. Subject 3-45N-m/deg ref BL ref HH act FF act AE-Early act AE-Late act de-adapt Percent Gait Cycle Figure 126: Subject 3 5 th trial. 148

163 Appendix C - RGR Trainer 2DOF Besides hip-hiking, another common secondary gait deviation occurring in the motion of the pelvis is circumduction with exaggerated pelvic rotation, as shown in Figure 5. Guiding pelvic rotation in order to affect this gait deviation requires the ability to generate moments in the horizontal plane. The design of a 2 DOF RGR Trainer, which can apply moments about pelvic obliquity and pelvic rotation, is presented. Requirements The device needs to be able to apply corrective moments to pelvic obliquity and pelvic rotation, while allowing close to free translations in the horizontal plane. As is the case with any impedance controlled device for human interaction, the inertia of the device should be kept to a minimum, in order to enhance the robot s ability to display the prescribed force fields. Considering the specific task at hand, it has been shown the inertia of the robot should be less than that of the actuated body part [47]. The body segment inertias can be found using equations in a NASA publication, based on the total body weight (TBW) [48]. For a light subject, these inertias can be as low as 22.9kg in vertical direction and 11.9kg in horizontal plane [49]. In that same work, the moment of inertia of the pelvis for a female 1% by weight was found to be kg-m 2. Static friction can cause the subject to lose balance. One study found that patients couldn t resist disturbances equal to 2% of their body weight [50]. Therefore, the maximum allowable static friction force in the horizontal plane was found to be 8.3N 149

164 [49]. That same source found the maximum desirable stiffness applied onto the body to be 4150N/m. Actuation System Design Concepts Concept 1 This concept design features two rigid links applying a moment at pelvic obliquity from behind, and two links applying moment at pelvic obliquity from above the patient. In this design concept the BWS system is integrated into the obliquity control parallelogram mechanism, the two linkages need to be designed to fully support a subject weighing 111kg, which amounts to 545N. The forces exerted onto the component as a result of moment application were calculated as well. Previous testing revealed that moments up to 60N-m have significant impact on simulated hip-hiking. Therefore, an upper-bound value of 100N-m was used for the moments applied to the pelvis. The moment-arm was found based on the 1 st percentile male s hip width. The moment arm was found to be 264mm, and therefore the necessary force transmitted by a single link necessary to impart the mentioned 100N-m moment was found to be 379N, putting the link in compression and subject to failure due to buckling. For the purpose of clearing the subject s upper body, the link was designed to approximate an arc, with an offset of about 140mm. A clearance of 165mm above the head of the 99 th percentile male was used. The link was subjected to a static test in tension of 924N (body weight support and applied moment) and a buckling test with a force of 379N (from moment application). The link s change in length due to compressive force (from moment application) amounted to just over 2mm. The result is a link which provides a safety factor of just above 2 for both modes, 150

165 and weighs in at about 1.1kg. With two such links positioned at either side of the 1% male pelvis, the resulting moment of inertia is approximately J= kg-m 2. This is an estimate of the minimum rotational inertia excluding the actuator and the control arm necessary to generate moments. Figure 127: Control link optimized for mass. This design s main drawback is the effect the vertical links and the overhead mechanism may have on the patient. This mechanism layout fails at becoming unobtrusive to the subject. The extra overhead clearance comes at the price of increased inertia, which is already significant. 151

166 Concept 2 Another option for applying a corrective moment to pelvic obliquity uses a shaft in torsion. Such a shaft should allow for unrestricted motion in the vertical and lateral directions (within certain limits), which points to a design with two flexible universal joints, as shown in (Figure 128). Figure 128: Concept 2 with torsion bar applying moments in pelvic obliquity and two push-rods in pelvic rotation, here pictured with two Copley linear actuators. Since the torsion bar is of fixed length, as it rotates, the links which apply moments in pelvic rotation must rotate with the torsion bar as well. Therefore, the system has a significant moment of inertia in pelvic obliquity. Of course, this system cannot provide body weight support. This can be accomplished by a separate overhead system. 152

167 Concept 3 This design consists of two parallel four-bar mechanisms, which support a cylindrical joint for pelvic obliquity, and a semi-circle supported by bearings, which operates at the pelvic rotation level with the remote center of rotation placed inside the subject s body (Figure 129). This design requires flexible transmission to deliver the driving moments to the two rotational DOF s. This reduces the mechanism s inertia (actuators are stationary), but the complexity of this design and necessity for high precision custommachined components make it a poor choice. Figure 129: Constant-radius arc guided by bearings used to place the center of rotation within the subject s body. Concept 4 The fourth concept features two triangular beams, which are constrained to rotate together. This creates a structure fixed in rotation but free to translate, for application of the moment to pelvic obliquity. 153

168 Figure 130: The gimbals at the hip joints allow for hip abduction/adduction and flexion/extension. Flexible transmission is required to apply moments at pelvic obliquity. This design relies heavily on the rigidity of the two triangular arms and the shaft connecting them, in order to apply the prescribed moment to pelvic obliquity. A well designed structure with high stiffness would result in high natural frequency, which is necessary to prevent control issues due to noisy force feedback signal. An offset axis of rotation at the base of the mechanism shifts the center of moment application in pelvic rotation to the inside of the body. Unfortunately, this design features significant moment of inertia in the controlled DOF, as well as complexity. Selected Concept This design consists of two planar manipulators, each composed of two linear actuators (Figure 131). Working in unison, these two manipulators can apply forces (vertical), moments (pelvic obliquity and pelvic rotation) or both onto the pelvic brace. In general, this mechanism cannot apply forces in the transverse direction (side to side) and forward- 154

169 back. As a result, the horizontal translations are non-actuated, but at the same time the mechanism can actively respond to environment s force input under zero-force-control, in order to minimize the interaction forces, hence it is backdrivable in horizontal translations. Figure 131: RGR Trainer 2DOF actuation concept. Each planar manipulator provides mounting for two linear actuators, which pivot about their center of housing, in order to minimize their moments of inertia. They are suspended on ball bearings to reduce friction as well. 155

170 Figure 132: Actuator mount detail design. Planar Manipulator Kinematics The actuation mechanism of the RTR Trainer 2DOF consists of four linear actuators, arranged in two closed-link mechanisms. The two mechanisms apply forces to the right and left sides of the pelvic brace. 156

171 Figure 133: Closed linkage mechanism. The following equations describe kinematics of the left-hand side closed link mechanism. The angle β L is found from the following equation: bl e a L L arccos 2* bl * e (C.1) Now using β, we can find the distance from the endpoint to the vertical axis: c b sin (C.2) L L L The angle of rotation is measured directly (α), so that we can now describe vector p, which locates the mechanism s endpoint with respect to the origin xyz. 157

172 p L c *sin ' c*cos b *cos (C.3) The position of the endpoint on the right-hand side is found in the same exact way, giving us two vectors, which describe the locations of P L and P R with respect to the two origins located on either side of the device. Next, we find the locations of these two points with respect to the default location of the subject in the device, as shown the figure below: Figure 134: Top view of the mechanism. Now we can employ translation in order to find the position vectors of points P L and P R with respect to the main reference frame (xyz): n PL PL ' C L 0 (C.4) 158

173 n P P ' C 0 R R R (C.5) The pelvic rotation Θ and pelvic obliquity φ angles are found using the above position vectors: n P P ' C 0 R R R (C.6) arctan P P L( z) R( z) PR ( z) PL ( z) PR ( y) PL ( y) 2 2 (C.7) Figure 135: Definition of pelvic obliquity Φ and pelvic rotation θ angles. Pelvic brace as viewed from behind. 159

174 Jacobian The two force vectors necessary to impart the desired moments Τ Θ and Τ φ onto the pelvis (Figure 136) are found as follows: arctan P P L( z) R( z) PR ( z) PL ( z) PR ( y) PL ( y) 2 2 (C.8) f R cos T sin sin T 2 Tz u cos T (C.9) Figure 136: Forces f L and f R necessary to produce desired net forces and moments. f 1 f v1 v2 f2 (C.10) Now substitute: 160

175 v i p p i (C.11) i and solve above equation for f i. f1 p2 p 2 f p p f (C.12) Thus we obtain the magnitudes of the force commands, shown in equation (12), which should be sent to the two actuators of a planar manipulator in order to produce the required force f, as is shown in Figure 137. The forces from the two manipulators (f R and f L ) produce the required torques, which are applied onto the pelvis through the pelvic interface. 161

176 Figure 137: Force f resolved into the component magnitudes and unit vectors. Device Structure Design RGR Trainer 2DOF Frame Requirements The primary purpose of the device s frame: - provide rigid support for the robot s manipulators such that forces can be safely and accurately applied onto a subject s pelvis. Secondary purposes: 162

177 - provide mounting for body weight support. - provide support for upper body (handle bar). Desirable features: - unrestricted access for physical therapist from the side to either leg of the subject. - ability to enter the device in wheel chair. - easy machine adjustment for different height patients. - simple, modular design for ease of installation on site. - low price. Frame Design In order to fulfill the requirements set forth, a simple structure was designed, which consists of rectangular cross section steel tubing, jointed by threaded fasteners (Figure 138 and Figure 139). The frame is wide enough for a wheel chair to enter the machine from the rear. The handlebar design allows for height adjustment as well as fore-aft adjustment. The robotic manipulators height is adjusted with a brake winch and cable pulley system, and is clamped into place with bolts, as shown in Figure 140. The handle bar tilt is adjustable when quick release clamps are loosened, while adding to the structural integrity of the frame when the clamps are tightened. Both adjustments can be performed by just one person and without need for tools thanks to use of knobs, indexing plunger and quick-release clamps (Figure 141 and Figure 142). 163

178 Figure 138: RGR Trainer 2DOF frame structure only. BWS beam is reinforced with steel cables. Figure 139: Side view of RGR Trainer 2DOF over treadmill. Manipulators are shown attached to pelvic interface worn by 99 th percentile female subject. 164

179 Figure 140: Manipulator attachment to frame. Figure 141: Handle bar tilt adjustment using quick release clamps and spring loaded plunger. 165

180 Figure 142: Handle bar height adjustment. Conclusion The RGR Trainer 2DOF design presented here expands the functionality of the RGR Trainer, by providing the ability to apply corrective torques to pelvic rotation. Abduction combined with exaggerated pelvic rotation is the second most common secondary gait deviation in control of the pelvis, after hip-hiking, and therefore in general it is desirable to be able to address this particular gait deviation in the future. 166

Ergonomic Handle for a 2DoF Robotic Hand Rehabilitation Device

Ergonomic Handle for a 2DoF Robotic Hand Rehabilitation Device Ergonomic Handle for a 2DoF Robotic Hand Rehabilitation Device Design Team Steven Adams, Kyle Hackmeister Lucas Johnson, Daniel Lau, Nicholas Pappas Design Advisor Prof. Constantinos Mavroidis Co-Advisors

More information

video Purpose Pathological Gait Objectives: Primary, Secondary and Compensatory Gait Deviations in CP AACPDM IC #3 1

video Purpose Pathological Gait Objectives: Primary, Secondary and Compensatory Gait Deviations in CP AACPDM IC #3 1 s in CP Disclosure Information AACPDM 71st Annual Meeting September 13-16, 2017 Speaker Names: Sylvia Ounpuu, MSc and Kristan Pierz, MD Differentiating Between, Secondary and Compensatory Mechanisms in

More information

Purpose. Outline. Angle definition. Objectives:

Purpose. Outline. Angle definition. Objectives: Disclosure Information AACPDM 69 th Annual Meeting October 21-24, 2015 Speaker Names: Sylvia Õunpuu, MSc and Kristan Pierz, MD Gait Analysis Data Interpretation: Understanding Kinematic Relationships Within

More information

EXPERIMENTAL STUDY OF EXOSKELETON FOR ANKLE AND KNEE JOINT

EXPERIMENTAL STUDY OF EXOSKELETON FOR ANKLE AND KNEE JOINT EXPERIMENTAL STUDY OF EXOSKELETON FOR ANKLE AND KNEE JOINT PROJECT REFERENCE NO. : 37S0925 COLLEGE : NEW HORIZON COLLEGE OF ENGINEERING, BANGALORE BRANCH : MECHANICAL ENGINEERING GUIDES : DR GANESHA PRASAD

More information

G-EOL. Discover the simplicity of gait therapy intended for daily use

G-EOL. Discover the simplicity of gait therapy intended for daily use G-EOL Discover the simplicity of gait therapy intended for daily use Reha Technology a passion for robotic-assisted gait therapy For over 10 years, Reha Technology has been successfully developing innovative,

More information

Motion Control of a Bipedal Walking Robot

Motion Control of a Bipedal Walking Robot Motion Control of a Bipedal Walking Robot Lai Wei Ying, Tang Howe Hing, Mohamed bin Hussein Faculty of Mechanical Engineering Universiti Teknologi Malaysia, 81310 UTM Skudai, Johor, Malaysia. Wylai2@live.my

More information

Normal and Abnormal Gait

Normal and Abnormal Gait Normal and Abnormal Gait Adrielle Fry, MD EvergreenHealth, Division of Sport and Spine University of Washington Board Review Course March 6, 2017 What are we going to cover? Definitions and key concepts

More information

C-Brace Orthotronic Mobility System

C-Brace Orthotronic Mobility System C-Brace Orthotronic Mobility System You ll always remember your first step Information for practitioners C-Brace Orthotics reinvented Until now, you and your patients with conditions like incomplete spinal

More information

In memory of Dr. Kevin P. Granata, my graduate advisor, who was killed protecting others on the morning of April 16, 2007.

In memory of Dr. Kevin P. Granata, my graduate advisor, who was killed protecting others on the morning of April 16, 2007. Acknowledgement In memory of Dr. Kevin P. Granata, my graduate advisor, who was killed protecting others on the morning of April 16, 2007. There are many others without whom I could not have completed

More information

Biomechanics and Models of Locomotion

Biomechanics and Models of Locomotion Physics-Based Models for People Tracking: Biomechanics and Models of Locomotion Marcus Brubaker 1 Leonid Sigal 1,2 David J Fleet 1 1 University of Toronto 2 Disney Research, Pittsburgh Biomechanics Biomechanics

More information

Humanoid Robots and biped locomotion. Contact: Egidio Falotico

Humanoid Robots and biped locomotion. Contact: Egidio Falotico Humanoid Robots and biped locomotion Contact: Egidio Falotico e.falotico@sssup.it Outline What is a Humanoid? Why Develop Humanoids? Challenges in Humanoid robotics Active vs Passive Locomotion Active

More information

WalkOn product range. Dynamic Ankle-Foot Orthoses. Information for specialist dealers

WalkOn product range. Dynamic Ankle-Foot Orthoses. Information for specialist dealers WalkOn product range Dynamic Ankle-Foot Orthoses Information for specialist dealers WalkOn Flex WalkOn WalkOn Trimable WalkOn Reaction WalkOn Reaction plus One range Many different applications The WalkOn

More information

Velocity-dependent reference trajectory generation for the LOPES gait training robot

Velocity-dependent reference trajectory generation for the LOPES gait training robot 2011 IEEE International Conference on Rehabilitation Robotics Rehab Week Zurich, ETH Zurich Science City, Switzerland, June 29 - July 1, 2011 Velocity-dependent reference trajectory generation for the

More information

Adaptation to Knee Flexion Torque Assistance in Double Support Phase

Adaptation to Knee Flexion Torque Assistance in Double Support Phase Adaptation to Knee Flexion Torque Assistance in Double Support Phase James S. Sulzer, Keith E. Gordon, T. George Hornby, Michael A. Peshkin and James L. Patton Abstract Studies have shown locomotor adaptation

More information

Complex movement patterns of a bipedal walk

Complex movement patterns of a bipedal walk 1 Complex movement patterns of a bipedal walk Objectives After completing this lesson, you will be able to: Describe the complex movement patterns of a bipedal walk. Describe the biomechanics of walking

More information

Kochi University of Technology Aca Study on Dynamic Analysis and Wea Title stem for Golf Swing Author(s) LI, Zhiwei Citation 高知工科大学, 博士論文. Date of 2015-03 issue URL http://hdl.handle.net/10173/1281 Rights

More information

Body Stabilization of PDW toward Humanoid Walking

Body Stabilization of PDW toward Humanoid Walking Body Stabilization of PDW toward Humanoid Walking Masaki Haruna, Masaki Ogino, Koh Hosoda, Minoru Asada Dept. of Adaptive Machine Systems, Osaka University, Suita, Osaka, 565-0871, Japan ABSTRACT Passive

More information

Coaching the Triple Jump Boo Schexnayder

Coaching the Triple Jump Boo Schexnayder I. Understanding the Event A. The Run and Its Purpose B. Hip Undulation and the Phases C. Making the Connection II. III. IV. The Approach Run A. Phases B. Technical Features 1. Posture 2. Progressive Body

More information

Gait Analyser. Description of Walking Performance

Gait Analyser. Description of Walking Performance Gait Analyser Description of Walking Performance This brochure will help you to understand clearly the parameters described in the report of the Gait Analyser, provide you with tips to implement the walking

More information

ABSTRACT. ambulatory aid by individuals with this type of permanent disability. This study

ABSTRACT. ambulatory aid by individuals with this type of permanent disability. This study ABSTRACT Title of dissertation: MODELING CRUTCH COMPENSATION OF HIP ABDUCTOR WEAKNESS AND PARALYSIS James Rocco Borrelli, Doctor of Philosophy, 2011 Dissertation directed by: Henry W. Haslach Jr. Department

More information

CHAPTER IV FINITE ELEMENT ANALYSIS OF THE KNEE JOINT WITHOUT A MEDICAL IMPLANT

CHAPTER IV FINITE ELEMENT ANALYSIS OF THE KNEE JOINT WITHOUT A MEDICAL IMPLANT 39 CHAPTER IV FINITE ELEMENT ANALYSIS OF THE KNEE JOINT WITHOUT A MEDICAL IMPLANT 4.1 Modeling in Biomechanics The human body, apart of all its other functions is a mechanical mechanism and a structure,

More information

An investigation of kinematic and kinetic variables for the description of prosthetic gait using the ENOCH system

An investigation of kinematic and kinetic variables for the description of prosthetic gait using the ENOCH system An investigation of kinematic and kinetic variables for the description of prosthetic gait using the ENOCH system K. OBERG and H. LANSHAMMAR* Amputee Training and Research Unit, University Hospital, Fack,

More information

GAIT RECOVERY IN HEALTHY SUBJECTS: PERTURBATIONS TO THE KNEE MOTION WITH A SMART KNEE BRACE. by Mehmet Temel

GAIT RECOVERY IN HEALTHY SUBJECTS: PERTURBATIONS TO THE KNEE MOTION WITH A SMART KNEE BRACE. by Mehmet Temel GAIT RECOVERY IN HEALTHY SUBJECTS: PERTURBATIONS TO THE KNEE MOTION WITH A SMART KNEE BRACE by Mehmet Temel A thesis submitted to the Faculty of the University of Delaware in partial fulfillment of the

More information

A CONTINOUS ROTARY ACTUATION MECHANISM FOR A POWERED HIP EXOSKELETON

A CONTINOUS ROTARY ACTUATION MECHANISM FOR A POWERED HIP EXOSKELETON University of Massachusetts Amherst ScholarWorks@UMass Amherst Masters Theses Dissertations and Theses 2015 A CONTINOUS ROTARY ACTUATION MECHANISM FOR A POWERED HIP EXOSKELETON Matthew C. Ryder University

More information

Servo-Assisted Lower-Body Exoskeleton With a True Running Gait

Servo-Assisted Lower-Body Exoskeleton With a True Running Gait Servo-Assisted Lower-Body Exoskeleton With a True Running Gait John Dick and Bruce Crapuchettes Applied Motion, Inc. 935 N. Indian Hill Blvd. Claremont, CA 91711 jdick@springwalker.com DARPA Workshop on

More information

Microprocessor Technology in Ankle Prosthetics

Microprocessor Technology in Ankle Prosthetics Microprocessor Technology in Ankle Prosthetics Arizona State University Dr. Thomas Sugar Former Students LTC Joseph Hitt, PhD Dr. Kevin Hollander Dr. Matthew Holgate Dr. Jeffrey Ward Mr. Alex Boehler Mr.

More information

Computer Aided Drafting, Design and Manufacturing Volume 26, Number 2, June 2016, Page 53. The design of exoskeleton lower limbs rehabilitation robot

Computer Aided Drafting, Design and Manufacturing Volume 26, Number 2, June 2016, Page 53. The design of exoskeleton lower limbs rehabilitation robot Computer Aided Drafting, Design and Manufacturing Volume 26, Number 2, June 2016, Page 53 CADDM The design of exoskeleton lower limbs rehabilitation robot Zhao Xiayun 1, Wang Zhengxing 2, Liu Zhengyu 1,3,

More information

Controlling Walking Behavior of Passive Dynamic Walker utilizing Passive Joint Compliance

Controlling Walking Behavior of Passive Dynamic Walker utilizing Passive Joint Compliance Controlling Walking Behavior of Passive Dynamic Walker utilizing Passive Joint Compliance Takashi TAKUMA, Koh HOSODA Department of Adaptive Machine Systems, Graduate School of Engineering Osaka University

More information

ZIPWAKE DYNAMIC TRIM CONTROL SYSTEM OUTLINE OF OPERATING PRINCIPLES BEHIND THE AUTOMATIC MOTION CONTROL FEATURES

ZIPWAKE DYNAMIC TRIM CONTROL SYSTEM OUTLINE OF OPERATING PRINCIPLES BEHIND THE AUTOMATIC MOTION CONTROL FEATURES ZIPWAKE DYNAMIC TRIM CONTROL SYSTEM OUTLINE OF OPERATING PRINCIPLES BEHIND THE AUTOMATIC MOTION CONTROL FEATURES TABLE OF CONTENTS 1 INTRODUCTION 3 2 SYSTEM COMPONENTS 3 3 PITCH AND ROLL ANGLES 4 4 AUTOMATIC

More information

DYNAMIC POSITIONING CONFERENCE October 7-8, New Applications. Dynamic Positioning for Heavy Lift Applications

DYNAMIC POSITIONING CONFERENCE October 7-8, New Applications. Dynamic Positioning for Heavy Lift Applications Return to Session Directory DYNAMIC POSITIONING CONFERENCE October 7-8, 2008 New Applications Dynamic Positioning for Heavy Lift Applications John Flint and Richard Stephens Converteam UK Ltd. (Rugby,

More information

ZMP Trajectory Generation for Reduced Trunk Motions of Biped Robots

ZMP Trajectory Generation for Reduced Trunk Motions of Biped Robots ZMP Trajectory Generation for Reduced Trunk Motions of Biped Robots Jong H. Park School of Mechanical Engineering Hanyang University Seoul, 33-79, Korea email:jong.park@ieee.org Yong K. Rhee School of

More information

intended velocity ( u k arm movements

intended velocity ( u k arm movements Fig. A Complete Brain-Machine Interface B Human Subjects Closed-Loop Simulator ensemble action potentials (n k ) ensemble action potentials (n k ) primary motor cortex simulated primary motor cortex neuroprosthetic

More information

Selective control of gait subtasks in robotic gait training: foot clearance support in stroke survivors with a powered exoskeleton

Selective control of gait subtasks in robotic gait training: foot clearance support in stroke survivors with a powered exoskeleton Koopman et al. Journal of NeuroEngineering and Rehabilitation 2013, 10:3 JOURNAL OF NEUROENGINEERING JNERAND REHABILITATION RESEARCH Open Access Selective control of gait subtasks in robotic gait training:

More information

Dynamically stepping over large obstacle utilizing PSO optimization in the B4LC system

Dynamically stepping over large obstacle utilizing PSO optimization in the B4LC system 1 Dynamically stepping over large obstacle utilizing PSO optimization in the B4LC system QI LIU, JIE ZHAO, KARSTEN BERNS Robotics Research Lab, University of Kaiserslautern, Kaiserslautern, 67655, Germany

More information

Simulation of the Hybtor Robot

Simulation of the Hybtor Robot Simulation of the Hybtor Robot Pekka Aarnio, Kari Koskinen and Sami Salmi Information and Computer Systems in Automation Helsinki University of Technology ABSTRACT A dynamic rigid body simulation model

More information

Modeling of Hydraulic Hose Paths

Modeling of Hydraulic Hose Paths Mechanical Engineering Conference Presentations, Papers, and Proceedings Mechanical Engineering 9-2002 Modeling of Hydraulic Hose Paths Kurt A. Chipperfield Iowa State University Judy M. Vance Iowa State

More information

Supplementary Figure S1

Supplementary Figure S1 Supplementary Figure S1: Anterior and posterior views of the marker set used in the running gait trials. Forty-six markers were attached to the subject (15 markers on each leg, 4 markers on each arm, and

More information

AN31E Application Note

AN31E Application Note Balancing Theory Aim of balancing How an unbalance evolves An unbalance exists when the principle mass axis of a rotating body, the so-called axis of inertia, does not coincide with the rotational axis.

More information

video Outline Pre-requisites of Typical Gait Case Studies Case 1 L5 Myelomeningocele Case 1 L5 Myelomeningocele

video Outline Pre-requisites of Typical Gait Case Studies Case 1 L5 Myelomeningocele Case 1 L5 Myelomeningocele Outline Evaluation of Orthosis Function in Children with Neuromuscular Disorders Using Motion Analysis Outcomes Terminology Methods Typically developing Case examples variety of pathologies Sylvia Õunpuu,

More information

+ t1 t2 moment-time curves

+ t1 t2 moment-time curves Part 6 - Angular Kinematics / Angular Impulse 1. While jumping over a hurdle, an athlete s hip angle was measured to be 2.41 radians. Within 0.15 seconds, the hurdler s hip angle changed to be 3.29 radians.

More information

Equine Cannon Angle System

Equine Cannon Angle System Equine Cannon System How to interpret the results December 2010 Page 1 of 14 Table of Contents Introduction... 3 The Sagittal Plane... 4 The Coronal Plane... 5 Results Format... 6 How to Interpret the

More information

Fail Operational Controls for an Independent Metering Valve

Fail Operational Controls for an Independent Metering Valve Group 14 - System Intergration and Safety Paper 14-3 465 Fail Operational Controls for an Independent Metering Valve Michael Rannow Eaton Corporation, 7945 Wallace Rd., Eden Prairie, MN, 55347, email:

More information

C-Brace Reimbursement Guide

C-Brace Reimbursement Guide Reimbursement Guide Information for practitioners and payers Product Information Effective September 24, 2018 The The is the first microprocessor stance and swing phase controlled orthosis (SSCO ). This

More information

A Bio-inspired Behavior Based Bipedal Locomotion Control B4LC Method for Bipedal Upslope Walking

A Bio-inspired Behavior Based Bipedal Locomotion Control B4LC Method for Bipedal Upslope Walking 1 A Bio-inspired Behavior Based Bipedal Locomotion Control B4LC Method for Bipedal Upslope Walking JIE ZHAO, QI LIU, STEFFEN SCHUETZ, and KARSTEN BERNS Robotics Research Lab, University of Kaiserslautern,

More information

The Effect of a Seven Week Exercise Program on Golf Swing Performance and Musculoskeletal Screening Scores

The Effect of a Seven Week Exercise Program on Golf Swing Performance and Musculoskeletal Screening Scores The Effect of a Seven Week Exercise Program on Golf Swing Performance and Musculoskeletal Screening Scores 2017 Mico Hannes Olivier Bachelor of Sport Science Faculty of Health Sciences and Medicine Bond

More information

Mechanical Knee Prescription Guide

Mechanical Knee Prescription Guide Mechanical Knee Prescription Guide CONTENTS K1 LOCKING KNEE 1 K1 - K2 BALANCE KNEE OFM1 2 BALANCE KNEE OFM2 3 K2 BALANCE KNEE OM8 4 TOTAL KNEE 1900 5 K2 - K3 OP2 KNEE 6 OP4 KNEE 7 OP5 KNEE 8 OHP3 KNEE

More information

RELIABILITY ASSESSMENT, STATIC AND DYNAMIC RESPONSE OF TRANSMISSION LINE TOWER: A COMPARATIVE STUDY

RELIABILITY ASSESSMENT, STATIC AND DYNAMIC RESPONSE OF TRANSMISSION LINE TOWER: A COMPARATIVE STUDY RELIABILITY ASSESSMENT, STATIC AND DYNAMIC RESPONSE OF TRANSMISSION LINE TOWER: A COMPARATIVE STUDY Yusuf Mansur Hashim M. Tech (Structural Engineering) Student, Sharda University, Greater Noida, (India)

More information

Selective control of a subtask of walking in a robotic gait trainer(lopes)

Selective control of a subtask of walking in a robotic gait trainer(lopes) Selective control of a subtask of walking in a robotic gait trainer(lopes) E. H.F. Van Asseldonk, R. Ekkelenkamp, Jan F. Veneman, F. C. T. Van der Helm, H. van der Kooij Abstract Robotic gait trainers

More information

Design of a double quadruped for the Tech United soccer robot

Design of a double quadruped for the Tech United soccer robot Design of a double quadruped for the Tech United soccer robot M.J. Naber (0571509) DCT report number: 2009.134 Master Open Space project Eindhoven, 21 December 2009 Supervisor dr.ir. P.C.J.N. Rosielle

More information

SCHEINWORKS Measuring and Analysis Systems by

SCHEINWORKS Measuring and Analysis Systems by Pressure Measurement Systems for standing and walking analysis Germany since 1879 Pressure Measurement Systems for standing and walking analysis Documentation of Gait image Stance Symmetry of all parameters

More information

Evaluation Of Impedance Control On A Powered Hip Exoskeleton

Evaluation Of Impedance Control On A Powered Hip Exoskeleton University of Massachusetts Amherst ScholarWorks@UMass Amherst Masters Theses Dissertations and Theses 2017 Evaluation Of Impedance Control On A Powered Hip Exoskeleton Punith condoor University of Massachusetts

More information

premise that interdependent body systems (e.g. musculoskeletal, motor, sensory, and cognitive

premise that interdependent body systems (e.g. musculoskeletal, motor, sensory, and cognitive APPENDIX 2 Motor Control Intervention Protocol The dynamic systems approach underlying motor control intervention is based on the premise that interdependent body systems (e.g. musculoskeletal, motor,

More information

Dynamic analysis and motion measurement of ski turns using inertial and force sensors

Dynamic analysis and motion measurement of ski turns using inertial and force sensors Available online at www.sciencedirect.com Procedia Engineering 6 ( 213 ) 355 36 6th Asia-Pacific Conference on Sports Technology Dynamic analysis and motion measurement of ski turns using inertial and

More information

WORKBOOK/MUSTANG. Featuring: The R82 Next Step Development Plan. mustang. R82 Education

WORKBOOK/MUSTANG. Featuring: The R82 Next Step Development Plan. mustang. R82 Education WORKBOOK/MUSTANG Featuring: The R82 Next Step Development Plan mustang R82 Education CLINICAL WORK BOOK/MUSTANG PAGE 2 PAGE 3 What is Mustang? Mustang is a highly adaptable walking aid for children and

More information

The Starting Point. Prosthetic Alignment in the Transtibial Amputee. Outline. COM Motion in the Coronal Plane

The Starting Point. Prosthetic Alignment in the Transtibial Amputee. Outline. COM Motion in the Coronal Plane Prosthetic Alignment in the Transtibial Amputee The Starting Point David C. Morgenroth, MD, Department of Rehabilitation Medicine University of Washington VAPSHCS Outline COM Motion in the Coronal Plane

More information

A NEW GOLF-SWING ROBOT MODEL UTILIZING SHAFT ELASTICITY

A NEW GOLF-SWING ROBOT MODEL UTILIZING SHAFT ELASTICITY Journal of Sound and Vibration (1998) 17(1), 17 31 Article No. sv981733 A NEW GOLF-SWING ROBOT MODEL UTILIZING SHAFT ELASTICITY S. SUZUKI Department of Mechanical System Engineering, Kitami Institute of

More information

Mobility Lab provides sensitive, valid and reliable outcome measures.

Mobility Lab provides sensitive, valid and reliable outcome measures. Mobility Lab provides sensitive, valid and reliable outcome measures. ith hundreds of universities and hospitals using this system worldwide, Mobility Lab is the most trusted wearable gait and balance

More information

Walking Simulator Mechanism

Walking Simulator Mechanism The Downtown Review Volume 2 Issue 2 Article 4 2015 Walking Simulator Mechanism Titus Lungu Cleveland State University Igor Tachynskyy Cleveland State University Omri Tayyara Cleveland State University

More information

Development of an end-effector to simulate the foot to ball interaction of an instep kick in soccer

Development of an end-effector to simulate the foot to ball interaction of an instep kick in soccer Available online at www.sciencedirect.com Procedia Engineering 34 (2012 ) 284 289 9 th Conference of the International Sports Engineering Association (ISEA) Development of an end-effector to simulate the

More information

ASSESMENT Introduction REPORTS Running Reports Walking Reports Written Report

ASSESMENT Introduction REPORTS Running Reports Walking Reports Written Report ASSESMENT REPORTS Introduction Left panel Avatar Playback Right Panel Patient Gait Parameters Report Tab Click on parameter to view avatar at that point in time 2 Introduction Software will compare gait

More information

Emergent walking stop using 3-D ZMP modification criteria map for humanoid robot

Emergent walking stop using 3-D ZMP modification criteria map for humanoid robot 2007 IEEE International Conference on Robotics and Automation Roma, Italy, 10-14 April 2007 ThC9.3 Emergent walking stop using 3-D ZMP modification criteria map for humanoid robot Tomohito Takubo, Takeshi

More information

GOLFER. The Golf Putting Robot

GOLFER. The Golf Putting Robot GOLFER The Golf Putting Robot Written By Wing Pan Yuen For EEL 5666 Intelligent Machines Design Laboratory December 05, 1999 Table of Contents Abstract Introduction Executive Summary Integrated System

More information

Passive Swing Assistive Exoskeletons for Motor-Incomplete Spinal Cord Injury Patients

Passive Swing Assistive Exoskeletons for Motor-Incomplete Spinal Cord Injury Patients 2007 IEEE International Conference on Robotics and Automation Roma, Italy, 10-14 April 2007 FrB8.4 Passive Swing Assistive Exoskeletons for Motor-Incomplete Spinal Cord Injury Patients Kalyan K Mankala,

More information

Decentralized Autonomous Control of a Myriapod Locomotion Robot

Decentralized Autonomous Control of a Myriapod Locomotion Robot Decentralized utonomous Control of a Myriapod Locomotion Robot hmet Onat Sabanci University, Turkey onat@sabanciuniv.edu Kazuo Tsuchiya Kyoto University, Japan tsuchiya@kuaero.kyoto-u.ac.jp Katsuyoshi

More information

The overarching aim of the work presented in this thesis was to assess and

The overarching aim of the work presented in this thesis was to assess and CHAPTER 7 EPILOGUE Chapter 7 The overarching aim of the work presented in this thesis was to assess and understand the effort for balance control in terms of the metabolic cost of walking in able-bodied

More information

Gripping rotary modules

Gripping rotary modules Gripping rotary modules Gripping rotary modules GRIPPING ROTARY MODULES Series Size Page Gripping rotary modules RP 314 RP 1212 318 RP 1216 322 RP 1520 326 RP 2120 330 RP 2128 334 RC 338 RC 1212 342 RC

More information

-Elastic strain energy (duty factor decreases at higher speeds). Higher forces act on feet. More tendon stretch. More energy stored in tendon.

-Elastic strain energy (duty factor decreases at higher speeds). Higher forces act on feet. More tendon stretch. More energy stored in tendon. As velocity increases ( ) (i.e. increasing Froude number v 2 / gl) the component of the energy cost of transport associated with: -Internal kinetic energy (limbs accelerated to higher angular velocity).

More information

Neurorehabil Neural Repair Oct 23. [Epub ahead of print]

Neurorehabil Neural Repair Oct 23. [Epub ahead of print] APPENDICE Neurorehabil Neural Repair. 2009 Oct 23. [Epub ahead of print] Segmental Muscle Vibration Improves Walking in Chronic Stroke Patients With Foot Drop: A Randomized Controlled Trial. Paoloni M,

More information

Inertial compensation for belt acceleration in an instrumented treadmill

Inertial compensation for belt acceleration in an instrumented treadmill Inertial compensation for belt acceleration in an instrumented treadmill Sandra K. Hnat, Antonie J. van den Bogert Department of Mechanical Engineering, Cleveland State University Cleveland, OH 44115,

More information

Active Orthosis for Ankle Articulation Pathologies

Active Orthosis for Ankle Articulation Pathologies Active Orthosis for Ankle Articulation Pathologies Carlos André Freitas Vasconcelos IST, Universidade Técnica de Lisboa Av. Rovisco Pais, 1049-001 Lisboa, Portugal Email: cafv@mail.com Abstract This work

More information

Mechanical Design of a Simple Bipedal Robot. Ming-fai Fong

Mechanical Design of a Simple Bipedal Robot. Ming-fai Fong Mechanical Design of a Simple Bipedal Robot by Ming-fai Fong Submitted to the Department of Mechanical Engineering in partial fulfillment of the requirements for the degree of Bachelor of Science in Mechanical

More information

Gait. Kinesiology RHS 341 Lecture 12 Dr. Einas Al-Eisa

Gait. Kinesiology RHS 341 Lecture 12 Dr. Einas Al-Eisa Gait Kinesiology RHS 341 Lecture 12 Dr. Einas Al-Eisa Definitions Locomotion = the act of moving from one place to the other Gait = the manner of walking Definitions Walking = a smooth, highly coordinated,

More information

Sample Solution for Problem 1.a

Sample Solution for Problem 1.a Sample Solution for Problem 1.a 1 Inverted Pendulum Model (IPM) 1.1 Equations of Motion and Ground Reaction Forces Figure 1: Scheme of the Inverted Pendulum Model (IPM). The equations of motion of this

More information

AN ISOLATED SMALL WIND TURBINE EMULATOR

AN ISOLATED SMALL WIND TURBINE EMULATOR AN ISOLATED SMALL WIND TURBINE EMULATOR Md. Arifujjaman Graduate Student Seminar: Master of Engineering Faculty of Engineering and Applied Science Memorial University of Newfoundland St. John s, NL, Canada

More information

ANNEXURE II. Consent Form

ANNEXURE II. Consent Form ANNEXURE II Consent Form I, voluntarily agree to participate in the research work entitled Gait Pattern in Post Stroke Hemiparetic Patients: Analysis and Correction. All my questions have been satisfactorily

More information

The Incremental Evolution of Gaits for Hexapod Robots

The Incremental Evolution of Gaits for Hexapod Robots The Incremental Evolution of Gaits for Hexapod Robots Abstract Gait control programs for hexapod robots are learned by incremental evolution. The first increment is used to learn the activations required

More information

Spider Robot for Motion with Quasistatic. Force Constraints

Spider Robot for Motion with Quasistatic. Force Constraints Spider Robot for Motion with Quasistatic Force Constraints Shraga Shoval, Elon Rimon and Amir Shapira Technion - Israel Institute of Technology - Haifa, Israel 32000. Abstract In quasistatic motions the

More information

C-Brace Reimbursement Guide

C-Brace Reimbursement Guide C-Brace Reimbursement Guide Information for practitioners and payers Product Information The C-Brace The C-Brace is the first microprocessor stance and swing phase controlled orthosis (SSCO). This highly

More information

Restraint Systems for Infants with Special Needs

Restraint Systems for Infants with Special Needs TEST METHOD 213.5 Restraint Systems for Infants with Special Needs Revised: Issued: May 2012R October 1997 (Ce document est aussi disponible en français) Table of Contents 1. Introduction... 1 2. Test

More information

Supplementary Information

Supplementary Information Supplementary Information Novel robotic interface to evaluate, enable, and train locomotion and balance after neuromotor disorders Nadia Dominici, Urs Keller, Heike Vallery, Lucia Friedli, Rubia van den

More information

Joint Torque Evaluation of Lower Limbs in Bicycle Pedaling

Joint Torque Evaluation of Lower Limbs in Bicycle Pedaling 11th conference of the International Sports Engineering Association, ISEA 216 Delft University of Technology; July 12 th Joint Torque Evaluation of Lower Limbs in Bicycle Pedaling Hiroki Yamazaki Akihiro

More information

Increasing ankle push-off work with a powered prosthesis does not necessarily reduce metabolic rate for transtibial amputees

Increasing ankle push-off work with a powered prosthesis does not necessarily reduce metabolic rate for transtibial amputees Supplementary Materials Increasing ankle push-off work with a powered prosthesis does not necessarily reduce metabolic rate for transtibial amputees Roberto E. Quesada, Joshua M. Caputo,, and Steven H.

More information

SECTION 4 - POSITIVE CASTING

SECTION 4 - POSITIVE CASTING 4-1 SECTION 4 - POSITIVE CASTING THE SHAPE OF THE SHELL IS DERIVED FROM THE SHAPE OF THE CAST Thermo-forming plastic for orthopedic intervention was originally developed at the University of California

More information

Optimization of an off-road bicycle with four-bar linkage rear suspension

Optimization of an off-road bicycle with four-bar linkage rear suspension Proceedings of MUSME 2008, the International Symposium on Multibody Systems and Mechatronics San Juan (Argentina), 8-12 April 2008 Paper n. 02-MUSME08 Optimization of an off-road bicycle with four-bar

More information

Toward a Human-like Biped Robot with Compliant Legs

Toward a Human-like Biped Robot with Compliant Legs Book Title Book Editors IOS Press, 2003 1 Toward a Human-like Biped Robot with Compliant Legs Fumiya Iida a,b,1, Yohei Minekawa a Juergen Rummel a and Andre Seyfarth a a Locomotion Laboratory, University

More information

Heat Engine. Reading: Appropriate sections for first, second law of thermodynamics, and PV diagrams.

Heat Engine. Reading: Appropriate sections for first, second law of thermodynamics, and PV diagrams. Heat Engine Equipment: Capstone, 2 large glass beakers (one for ice water, the other for boiling water), temperature sensor, pressure sensor, rotary motion sensor, meter stick, calipers, set of weights,

More information

INTERACTION OF STEP LENGTH AND STEP RATE DURING SPRINT RUNNING

INTERACTION OF STEP LENGTH AND STEP RATE DURING SPRINT RUNNING INTERACTION OF STEP LENGTH AND STEP RATE DURING SPRINT RUNNING Joseph P. Hunter 1, Robert N. Marshall 1,, and Peter J. McNair 3 1 Department of Sport and Exercise Science, The University of Auckland, Auckland,

More information

Ranger Walking Initiation Stephanie Schneider 5/15/2012 Final Report for Cornell Ranger Research

Ranger Walking Initiation Stephanie Schneider 5/15/2012 Final Report for Cornell Ranger Research 1 Ranger Walking Initiation Stephanie Schneider sns74@cornell.edu 5/15/2012 Final Report for Cornell Ranger Research Abstract I joined the Biorobotics Lab this semester to gain experience with an application

More information

HRC adjustable pneumatic swing-phase control knee

HRC adjustable pneumatic swing-phase control knee HRC adjustable pneumatic swing-phase control knee S. NAKAMURA and S. SAWAMURA Hyogo Rehabilitation Centre, Kobe, Japan Abstract Since 1972 the Hyogo Rehabilitation Centre has been developing a variable-resistance-type

More information

Evolving Gaits for the Lynxmotion Hexapod II Robot

Evolving Gaits for the Lynxmotion Hexapod II Robot Evolving Gaits for the Lynxmotion Hexapod II Robot DAVID TOTH Computer Science, Worcester Polytechnic Institute Worcester, MA 01609-2280, USA toth@cs.wpi.edu, http://www.cs.wpi.edu/~toth and GARY PARKER

More information

ME 8843-Advanced Mechatronics. Project Proposal-Automatic Bike Transmission

ME 8843-Advanced Mechatronics. Project Proposal-Automatic Bike Transmission ME 8843-Advanced Mechatronics Project Proposal-Automatic Bike Transmission 1/21/09 Razid Ahmad Brandon Borm Todd Sifleet Project Proposal: Our goal for the semester long project is to create and automatic

More information

Serve the only stroke in which the player has full control over its outcome. Bahamonde (2000) The higher the velocity, the smaller the margin of

Serve the only stroke in which the player has full control over its outcome. Bahamonde (2000) The higher the velocity, the smaller the margin of Lower Extremity Performance of Tennis Serve Reporter: Chin-Fu Hsu Adviser: Lin-Hwa Wang OUTLINE Introduction Kinetic Chain Serve Types Lower Extremity Movement Summary Future Work INTRODUCTION Serve the

More information

Analysis of ankle kinetics and energy consumption with an advanced microprocessor controlled ankle foot prosthesis.

Analysis of ankle kinetics and energy consumption with an advanced microprocessor controlled ankle foot prosthesis. Analysis of ankle kinetics and energy consumption with an advanced microprocessor controlled ankle foot prosthesis. D.Moser, N.Stech, J.McCarthy, G.Harris, S.Zahedi, A.McDougall Summary This study reports

More information

Running injuries - what are the most important factors

Running injuries - what are the most important factors Created as a free resource by Clinical Edge Based on Physio Edge podcast 59 with Greg Lehman, Tom Goom and Dr Christian Barton Get your free trial of online Physio education at Why do runners get injured?

More information

Improvement of the Cheetah Locomotion Control

Improvement of the Cheetah Locomotion Control Improvement of the Cheetah Locomotion Control Master Project - Midterm Presentation 3 rd November 2009 Student : Supervisor : Alexander Sproewitz Professor : Auke Jan Ijspeert Presentation of the Cheetah

More information

Putting Report Details: Key and Diagrams: This section provides a visual diagram of the. information is saved in the client s database

Putting Report Details: Key and Diagrams: This section provides a visual diagram of the. information is saved in the client s database Quintic Putting Report Information Guide Putting Report Details: Enter personal details of the client or individual who is being analysed; name, email address, date, mass, height and handicap. This information

More information

GOLF SPECIFIC DYNAMIC WARM UP

GOLF SPECIFIC DYNAMIC WARM UP GOLF SPECIFIC DYNAMIC WARM UP Golf-related injury is common. The three most common areas injured include: 1. The back 2. The wrists, and 3. The elbows. A golf-specific dynamic warm-up is recommended by

More information

Denny Wells, Jacqueline Alderson, Kane Middleton and Cyril Donnelly

Denny Wells, Jacqueline Alderson, Kane Middleton and Cyril Donnelly 11:45 am-12:00 pm Denny Wells. Assessing the accuracy of inverse kinematics in OpenSim to estimate elbow flexionextension during cricket bowling: Maintaining the rigid linked assumption. (201) ASSESSING

More information

Artifacts Due to Filtering Mismatch in Drop Landing Moment Data

Artifacts Due to Filtering Mismatch in Drop Landing Moment Data Camenga et al. UW-L Journal of Undergraduate Research XVI (213) Artifacts Due to Filtering Mismatch in Drop Landing Moment Data Elizabeth T. Camenga, Casey J. Rutten, Brendan D. Gould, Jillian T. Asmus,

More information

PERCEPTIVE ROBOT MOVING IN 3D WORLD. D.E- Okhotsimsky, A.K. Platonov USSR

PERCEPTIVE ROBOT MOVING IN 3D WORLD. D.E- Okhotsimsky, A.K. Platonov USSR PERCEPTIVE ROBOT MOVING IN 3D WORLD D.E- Okhotsimsky, A.K. Platonov USSR Abstract. This paper reflects the state of development of multilevel control algorithms for a six-legged mobile robot. The robot

More information