Development and control of a one-wheel telescopic active cane

Size: px

Start display at page:

Download "Development and control of a one-wheel telescopic active cane"

Violet Ellis
5 years ago
Views:

1 24 5th IEEE RAS & EMBS Internationa Conference on Biomedica Robotics and Biomechatronics (BioRob) August 2-5, 24. São Pauo, Brazi Deveopment and contro of a one-whee teescopic active cane Ragou Ady,2,3, Wae Bachta,2,3 and Phiippe Bidaud,2 Abstract Popuation ageing cas for innovative soutions to increase daiy iving autonomy. Since peope autonomy reies on their mobiity capabiities, severa robotized waking aids i.e. wakers and canes, mainy incuding navigation functions, have been deveoped. The existing robotized canes generay consist in staticay stabe mobie patforms equipped with a rod and a hande. This design aters the basic characteristics of origina canes i.e. their weight-ightness and compactness. In this paper, a one-whee teescopic active cane coser to the origina cane concept is presented. Its contro aw and synchronization with the waking cyce is aso given aong with experimenta resuts. I. INTRODUCTION Popuation ageing cas for the deveopment of new technoogies to hep enhancing the edery s quaity of ife. The main chaenge consists in increasing peope s autonomy in their daiy-ife activities. A first step towards increasing autonomy consists in improving the edery s mobiity so they recover the abiity to interact with their environment. Among mobiity aids, wakers and canes are generay preferred to wheechairs when their use is sti compatibe with the subject s heath condition. Indeed waking heps the edery maintaining a correct postura equiibrium and a good heath condition [5]. In this context, severa robotized wakers [3], [4], [8], [] and canes [2], [4], [5] have been deveoped. Their main goa is to provide more assistance and user friendiness than conventiona aids. The ast point is critica since it has been shown that using assistive devices increases attention demand among the edery [6] threatening their postura stabiity. To meet these objectives, robotized aids are controed using different strategies. Some are programmed to adapt their behavior with respect to the subject s intention or hep tracking a predefined trajectory, whie others are intended to secure peope s postura stabiity. Robotized assistants are currenty composed of conventiona assistants ike devices mounted on staticay stabe mobie robots i.e. patforms with at east three whees. Whie this design approach does not consideraby increase the size of robotized wakers when compared to conventiona ones, it aters the main characteristics of canes. Indeed, the incuded mobie patforms have arger support patforms when compared to Sorbonne Universités, UPMC Univ. Paris 6, Institut des Systèmes Inteigents et de Robotique, F-755, Paris, France. 2 CNRS, UMR 7222, Institut des Systèmes Inteigents et de Robotique, F-755, Paris, France. 3 INSERM, U5, Institut des Systèmes Inteigents et de Robotique, Equipe Agathe, F-755, Paris, France. This work is accompished within the aboratory of exceence SMART supported by French state funds managed by the ANR within the Investissements d Avenir program (ANR--IDEX-4-2). the dista tip of conventiona canes. This resuts in devices not ooking ike canes. The base of support size of robotized canes coud not be reduced without increasing faing over risks. One ongoing project in our aboratory consist in deveopping a one-whee teescopic active cane. This design aows to stay coser to conventiona canes characteristics. Indeed with one whee, it is easier to design a thin and compact cane reducing the support surface to the ony contact between the whee and the foor. In this paper the current prototype design is presented. The issue of synchronizing the active cane with a straightforward gait is tacked as we. The seque of the paper is organized as foowing. In section II, the deveoped cane is described. In section III a conventiona cane assisted waking is anayzed in order to derive the contro strategy for the deveoped active cane. This strategy is given in section IV. Experimenta resuts are then presented in section V. Concusions and future works are finay discussed. II. ACTIVE CANE PROTOTYPE A one-whee teescopic active cane prototype weighing 5,7 kg has been deveoped (Fig.). In this section, its mechanica architecture, its intended use and its embedded contro eectronics are described. inear axis motorized whee hande nut screw motor+encoder + brake battery ocation motor whee Fig.. Cane prototype ((eft) a schematic of the cane kinematics (midde) a CAD view (right) the current prototype) /6/4/$3. 24 IEEE 46

A. Mechanica architecture The robotized cane is composed of a inear actuated axis and a motorized whee (Fig.). The inear axis is actuated using a Maxon EC-Max 3 brushess motor.

Its nomina veocity and force are respectivey.33 m.s and 5 N. A faisafe brake is equipping this axis to maintain a constant ength when the power is off.

2 A. Mechanica architecture The robotized cane is composed of a inear actuated axis and a motorized whee (Fig.). The inear axis is actuated using a Maxon EC-Max 3 brushess motor. The rotary to inear motion conversion is ensured through a ba screw mechanism. The inear axis dispacement range is of.5 m admitting a cane ength variation from.8 m to.95 m. Its nomina veocity and force are respectivey.33 m.s and 5 N. A faisafe brake is equipping this axis to maintain a constant ength when the power is off. The whee is actuated using a bet-pueys mechanism powered by an ECfat motor equipped with a Maxon GP52-2 gear. The whee nomina veocity and torque are respectivey 23 rad.s and 4.6 N.m. Both motors are equipped with incrementa optica encoders. As the whee diameter is equa to. m, the nomina cane inear tip veocity and tangentia force are respectivey. m.s and 92 N. B. Intended use The deveoped active cane is intended to be hed ike a conventiona cane. The motorized rod and whee aow the cane to move without appying forces on the subject s hand. Unike conventiona canes, the proposed active device is not expected to be ifted and put on the foor at each stride. Instead, it shoud track the subject s forward gait and stop moving to provide mechanica support when necessary. The gait tracking is achieved by the means of a wireess sensor equipping the subject. When the subject takes a turn, he/she just has to rotate the active cane aong its axis and continue waking. C. Contro eectronics The motors actuating the teescopic axis and the whee are driven using two Soo-Whiste servo drives set in veocity contro mode. The reference veocities are sent to the servo drives using a mbed LPC768 micro controer through a CAN. Reference veocities are computed based on the measurements of the cane state and on two Inertia Measurement Unit (IMU) signas. One IMU is attached to the cane whie the second is equipping the cane user. The two IMUs are connected to an Arduino Due board using respectivey a seria and buetooth communication protocos. The Arduino board sends the acquired data to the mbed LPC768 through a seria. The two servo drives, the mbed and the arduino board are powered using two LiPo batteries. Figure 2 iustrates the cane eectronic wiring diagram. IMU IMU Seria Fig. 2. Buetooth Shied Seria Arduino Due Seria MBED LPC 768 Cane eectronic wiring diagram CAN Servo Drive Rod Servo Drive Whee III. CONVENTIONAL CANE ASSISTED-WALKING In this section, an anaysis of straightforward conventiona passive cane assisted gait is given. This anaysis is achieved to determine the strategy for synchronizing the active cane motion with the gait cyce during forward waking. First, the cane assisted gait structure is recaed. Then measurements of both thigh and passive cane anges enabes to seect a contro strategy for the active cane. A. Cane assisted gait structure The cycica properties of human straightforward gait aows considering it as a succession of strides. During a stride, each eg aternates between a stance and a swing phase. The transition between the two phases occurs after a doube stance during which both feet touch the foor. When a eg is impaired, a cane may be prescribed to assist the gait. A cane may be used in a controatera or ipsiatera way. Without oss of generaity, a controatera use of the cane is considered hereafter [3], [7]. In a controatera use (Fig. 3), the cane is hed on the opposite side of the weakest imb enabing a reduced oad on both the affected imb and hip [6]. Aside from enhancing the postura stabiity by increasing the base of support, a cane provides breaking and propusive forces to support the subject s weakest eg [2], [9]. The breaking and propusive forces are needed respectivey during the hee strike and the push off of the weakest eg. Figure 3 depicts a cane-assited stride. It begins with the pushoff performed by the weak eg assisted by the cane. Then, the weak eg and the cane swing together whie the other eg remains in contact with the foor. The cane supports the weak eg during the hee strike. The sound eg swings, whie the weak eg foot and the cane tip remain motioness (d), unti the end of the cyce (e). B. Cane assisted gait anaysis The previous description of the assisted gait structure shows a strong correation between the passive cane and the weak eg dispacements. Foowing that, experimenta resuts confirming this observation are given. ) Experimenta Setup: Four heathy subjects equipped with a knee brace and a modified soe have participated to the experiments. The equipment is intended to simuate waking troubes []. They were first asked to wak during five minutes to get used to the experimenta conditions. Secondy, the subjects were asked to wak with the hep of a passive cane during 5 minutes. A CodaMotion system composed of three infrared cameras has been used to capture their motion. To this purpose, both subjects imbs and cane have been equipped with active infrared markers (see Fig. 4). 2) Resuts: The thigh and cane orientations in the sagitta pane have been computed using the markers positions captured during the cane assisted waking. These anges for the four subjects during a straightforward gait are potted in Fig. 5. The pots show a good superimposition between the two anges. A synchrony and an amost simiar range are observed between the orientation of the cane and the weak 462

3 (d) (e) Fig. 3. Cane assisted stride first described. The active cane modeing and the contro aw are then presented. Fig. 4. Codamotion camera Infrared markers Cane with markers Motion capture markers positions on the subjects and the cane eg thigh during the cane assisted wak. This correation wi be used in the next section to eaborate a contro strategy for the active cane (d) Fig. 5. Comparison between cane (back ine) and impaired eg thigh (red ine) orientation during cane assisted waking IV. CONTROL SCHEME In the foowing section, the strategy adopted to synchronize the cane motion and the straightforward gait is A. Strategy Given the resuts obtained in the ast section, we propose to contro the cane orientation to track the weakest eg ange. This aows to assist this eg when necessary. To this aim, an IMU providing the ange that shoud be tracked by the active cane is strapped to the subject s thigh (see Fig. ). Using this strategy eads to the active cane behavior depicted in Fig. 6. The cane activey tracks the orientation of the weak eg during its swing phase. During the doube stance phase, the active cane does not move since the thigh ange does not vary significanty. At the beginning of the sound eg swing phase, the anges of the cane and the weak eg s thigh are equa. During the remainder of this phase, as the subject s hand is moving forward with the trunk, the cane orientation and the thigh s ange vary in the same way. This resuts in a motioness cane tip and rod. Again during the doube stance phase, the cane remains at rest unti the beginning of the weak eg swing. Note that this strategy is in accordance with [7] where the authors take advantage of the synergy between the cane and the egs motion to contro a HAL Exoskeeton. B. Modeing and contro aw In this subsection the thigh ange tracking contro scheme is described. Before giving the contro aw, the differentia kinematic mode of the active cane is first derived. This mode gives the reationship between the cane anguar veocity θ and the generaized veocities d and (Fig. 7). d and denote the veocities of respectivey the cane tip and the teescopic inear axis. The cane orientation θ, d and are reated as foows: sinθ = d Differentiating this equation gives: θ = () d cos(θ) d 2 cos(θ) (2) 463

(d) (e) (f) Fig. 6. Active Cane assisted stride θ θ d + _ K ( ) ḋ u = J + h θ IMU Fig. 8. Cane contro scheme Fig. 7.

This avoids odometry errors that arise during ong distance motion []. To foow the subject s gait, the cane orientation shoud be servoed to the impaired eg s one denoted θ d. As depicted in Fig.

d and r are sent as reference veocities to the motors drivers (where r is the whee radius).

4 (d) (e) (f) Fig. 6. Active Cane assisted stride θ θ d + _ K ( ) ḋ u = J + h θ IMU Fig. 8. Cane contro scheme Fig. 7. Cane mode, ength of the cane, h height of the hande, d handewhee foor distance This equation can be rewritten as: [ ] ḋ θ = J (3) [ ] where : J = cos(θ) d 2 cos(θ) is the Jacobian matrix of the active cane. Since d = sin(θ), the Jacobian matrix has been rewritten as foowing: [ J = (4) d cos(θ) ] tan(θ) This way, the Jacobian wi not depend on d and thus on the motorized whee encoder information. This avoids odometry errors that arise during ong distance motion []. To foow the subject s gait, the cane orientation shoud be servoed to the impaired eg s one denoted θ d. As depicted in Fig. 8, the contro aw writes: [ ḋ ] = K cosθ +sin 2 θ sinθcosθ +sin 2 θ } {{ } J + (θ d θ) (5) where : K is a proportiona contro gain. J + is the pseudoinverse of the Jacobian matrix J. d and r are sent as reference veocities to the motors drivers (where r is the whee radius). The veocity servoing oop is considered instantaneous since it is much faster than the cane orientation contro oop. V. RESULTS Before describing the synchronization of the active cane motion with a straightforward gait, the basic feedback contro properties of the impemented contro aw are given. A. Preiminary experiments According to the experiments presented in section III, the cane orientation range is [ 25,25 ]. To achieve an experimenta evauation of the active cane contro aw, an orientation reference trajectory composed of 3 and 3 magnitude steps is considered. During this experiment, the cane hande is kept at a stationary position by a subject. Figure 9 shows the cane response to the above described reference trajectory. The rod and the cane tip veocities are aso given. The obtained tracking performances are satisfying with an amost unit gain and an about.4 second rise time. The cane orientation response to the first 3 magnitude step is achieved by a backward dispacement of the tip (negative speed) and an extension of the teescopic rod (positive speed). As for the second reference change (from 3 to 3 ), the cane tip moves forward. The rod 464

4 2 B. Synchronization with the gait cyce Θ (deg) 2. 4 5 Rod speed Cane tip speed Cane IMU.5.5. 5.5.5 5 Fig. 9. Cane response to a reference orientation trajectory composed of 3, 3 and 3 steps.

5 4 2 B. Synchronization with the gait cyce Θ (deg) Rod speed Cane tip speed Cane IMU Fig. 9. Cane response to a reference orientation trajectory composed of 3, 3 and 3 steps. θ d in green and θ in back Thigh wireess IMU shortens unti the cane is vertica and then extends. Fig Θ (deg) Variations of Θ 5 5 Rod speed Cane tip speed Disturbance rejection experiment for a reference orientation of A second experiment has been conducted. In this experiment, the cane reference orientation is set to 3 degrees. The subject hoding the cane hande is asked to change his hand position to induce four disturbances. The active cane is supposed to move in order to keep its desired orientation. The behavior of the cane during this experiment is shown in Fig.. One can see that when a disturbance occurs due to the subject s hand motion, the cane activey recovers its desired orientation as expected. Fig.. A subject equipped with a wireess IMU on his thigh and hoding the active cane Fig. 2. Anges (deg) Anges (deg) Cane foowing experiments, θ d in green, θ in back In the foowing, the resuts obtained using the strategy described in section IV are given and discussed. A wireess IMU is attached to the subject s thigh (Fig. ). The experiment begins with the subject standing sti and keeping his feet parae. The cane is hed verticay. The 465

6 subject is then asked to begin waking by swinging first the eg equipped with the IMU. Figure 2 shows the resuts of two waking experiments with two different stride engths of amost.87 m (Fig. 2 a) and.96m (Fig. 2 b). The cane has achieved a quite good tracking of the weak eg ange with respectivey a RMSE of 6 and 7. The cane behavior during one stride is given in Fig. 3. The active cane begins by tracking the orientation of the weak eg during its swing phase (Fig. 6 a-c). From the weak eg hee-strike unti the end of its stance phase, the cane rod and tip are motioness (Fig. 6 c-f), aowing the subject to use it ike a passive conventiona cane. Forces can thus be appied safey on the cane. This behavior is expained by the subject s hand motion resuting in a rotation of the cane which equa to the thigh ange rotation. For more safety, the rod and tip veocities may be set to zero at the beginning of the doube support phase detected by the IMU attached to the subject s thigh Anges (deg) Rod speed. 2 3 Fig Cane tip speed 2 3 Cane behavior during a stride, θ d in green, θ in back VI. CONCLUSIONS In this paper, the deveopment of a one-whee teescopic active cane has been introduced. This cane keeps the main characteristics of a conventiona cane, namey its compactness and reduced tip size. The synchronization of the active cane motion with a straightforward gait has aso been achieved. To this purpose the cane orientation has been servoed to the ange of the weakest eg thigh. Promising experimenta resuts have been obtained. Future work incudes two main topics. The first one consists in the assessment of the active cane benefits. To this aim, a comparison of the forces appied on the active cane have to be measured and compared to those obtained using a conventiona cane. Besides, the use easiness has to be investigated, through experiments invoving subjects with gait impairments. For that purpose, the safety of the cane wi be enhanced to cope with gait pattern disruption or fas. The second topic is reated to contro issues. Indeed, in the present work, ony straightforward waking is considered. The cane contro shoud hande other situations e.g. when the waking direction changes or the more chaenging stairs cimbing. REFERENCES [] Marko Ackermann and Werner Schiehen. Dynamic Anaysis of Human Gait Disorder and Metaboica Cost Estimation. Archive of Appied Mechanics, 75(-2): , May 26. [2] R. Ady, W. Bachta, and P. Bidaud. Anaysis of cane-assisted waking through noninear optimization. In Robotics and Automation (ICRA), 23 IEEE Internationa Conference on, pages , May 23. [3] Steven Dubowsky, Frank Gnot, Sara Godding, Hisamitsu Kozono, Adam Skwersky, Haoyong Yu, and Long Shen Yu. Pamm - a robotic aid to the edery for mobiity assistance and monitoring: A hepinghand for the edery. In ICRA, pages , 2. [4] B. Graf, M. Hans, and D. S. Rof. Care-o-bot ii deveopment of a next generation robotic home assistant. Autonomous Robots, 6:93 25, 24. [5] a. a. Hakim, J. D. Curb, H. Petrovitch, B. L. Rodriguez, K. Yano, G. W. Ross, L. R. White, and R. D. Abbott. Effects of Waking on Coronary Heart Disease in Edery Men : The Honouu Heart Program. Circuation, ():9 3, Juy 999. [6] Brian E. Maki Hamid Bateni. Assistive devices for baance and mobiity: Benefits, demands, and adverse consequences. Archives of Physica Medicine and Rehabiitation, 86():34 45, 25. [7] Modar Hassan, Hideki Kadone, Kenji Suzuki, and Yoshiyuki Sankai. Exoskeeton robot contro based on cane and body joint synergies. 22 IEEE/RSJ Internationa Conference on Inteigent Robots and Systems, pages 69 64, October 22. [8] Y. Hirata, S. Komatsuda, and K. Kosuge. Fa prevention contro of passive inteigent waker based on human mode. In Inteigent Robots and Systems, 28. IROS 28. IEEE/RSJ Internationa Conference on, pages , sept. 28. [9] L.Bennett, M.P. Murray, E. Murphy, and T.Sowe. Locomotion assistance through cane impuse. Pubic Heath, pages 38 47, 978. [] Kooktae Lee, Woojin Chung, and Kwanghyun Yoo. Kinematic parameter caibration of a car-ike mobie robot to improve odometry accuracy. Mechatronics, 2(5): , August 2. [] Aaron Morris, Raghavendra Donamukkaa, Anuj Kapuria, Aaron Steinfed, Judy Matthews, Jackie Dunbar-Jacobs, and Sebastian Thrun. A robotic waker that provides guidance. In Proceedings of IEEE Internationa Conference on Robotics and Automation (ICRA 3, pages 25 3, 23. [2] S. Nakagawa, D. Pei, Jian Huang, K. Sekiyama, and T. Fukuda. Contro of inteigent cane robot considering usage of ordinary cane. In RO-MAN, 23 IEEE, pages , Aug 23. [3] M Simic, K L Benne, M a Hunt, T V Wrigey, and R S Hinman. Contraatera cane use and knee joint oad in peope with media knee osteoarthritis: the effect of varying body weight support. Osteoarthritis and cartiage / OARS, Osteoarthritis Research Society, 9():33 7, November 2. [4] S. Suzuki, Y. Hirata, and K. Kosuge. Deveopment of inteigent passive cane controed by servo brakes. In IEEE Internationa Symposium on Robot and Human Interactive Communication, pages 97 2, 29. [5] K. Wakita, J. Huang, Pei Di, K. Sekiyama, and T. Fukuda. Humanwaking-intention-based motion contro of an omnidirectiona-type cane robot. IEEE/ASME Transactions on Mechatronics, PP(99): 2, 2. [6] R. Wemon, K. Pezzio, G. Eichhorn, W. Lockhart, and J. Morris. Changes in dua-task voice reaction time among eders who use assistive devices. Journa of geriatric physica therapy, 29(2):74 8, 2. [7] M. Whitte. Gait anaysis: an introduction. Butterworth Heinemann. Butterworth-Heinemann,

2014 John Wiley & Sons, Inc. All rights reserved.

2014 John Wiley & Sons, Inc. All rights reserved. 2014 John Wiey & Sons, Inc. A rights reserved. 15-2 Energy in Simpe Harmonic Motion 15-2 Energy in Simpe Harmonic Motion Write the functions for kinetic and potentia energy: Eq. (15-18)18 Their sum is