The th International onference on Design Engineering and Science, IDES 7 Aachen, Germany, Setember 7-9, 7 A Study on Autonomous Oeration System of aisson Shovels in High Air Pressure and Narrow Underground Sace (Demonstration of Trajectory Tracking ontrol) Naoto NEGISHI *, Toshitaka TSUNEKI *, Koki KIKUHI*, Toshihiro KONDO*, Tetsuya KOYO*, Akira KAMEI* and Keigo HAYAKAWA* * The Deartment of Advance Robotics, hiba Institute of Technology -7- Tsudanuma, Narashino, hiba 75-6, JAPAN s68yu@s.chibakoudai.j, s68wt@s.chibakoudai.j, kikut@ieee.or.j * Oriental Shiraishi or. 5-6-5 Toyosu, Koto-ku, Tokyo 5-6, JAPAN Abstract In this aer, we develo an unmanned construction system using autonomous caisson shovels with a characteristic joint configuration. The caisson shovel is a link maniulator with five degrees of freedom (DOFs) and travels on a rail track of a working chamber ceiling in a caisson box. Here, to measure the joint variables, we first installed three laser sensors for traveling on the rail track, duming, and extending the boom unit, an encoder for yawing the boom unit, and a air of ultrasonic flow sensors for itching the bucket. We then reroduced the trajectory of the caisson shovel during a soil mountain removal task by a skilled human oerator using a roortional-derivative (PD) control and assessed the reroducibility and the motion characteristics of the system. The exerimental results revealed that although the ultrasonic flow sensor had hysteresis and additionally the caisson shovel was subject to large inertia and friction, the tracking accuracy of this system was sufficient for simle oeration. Keywords: unmanned construction, neumatic caisson method, caisson shovel, trajectory tracking Introduction A neumatic caisson method (PM) [], [] is a neumatic rocess for constructing bridge foundations, underground retention basins, etc. The caisson shovel, which is mounted on a rail track of a working chamber ceiling, excavates the ground and immerses the steel reinforced concrete caisson box vertically into the underground (Figure ). The inside of the working chamber is sealed and kees high air ressure deending on the underground deth to avoid the intrusion of the groundwater. From this reason, since the excavation surface is not submerged, construction is ossible even in water. The maximum deth and ressure are aroximately 7m and.7mpa, resectively. The caisson shovel, a link maniulator with five degrees of freedom (DOFs), is assembled and disassembled through the material shaft connected to the working chamber and is teleoerated from an oeration room above the ground through the images of the cameras mounted on the ceiling and the shovel. Although the teleoerated caisson shovel has solved many roblems for safety, several issues and roblems remain. () There are insufficient shovel oerators. In a large construction area such as 7m*7m, for examle, caisson shovels would require human oerators. () There are also insufficient skilled oerators. Learning the required skill is very time consuming. () Even routine simle tasks such as soil mountain removal must be erformed by skilled oerators. () Since the D camera image for teleoeration limits the field of view and loses distance information, the excavation efficiency deteriorates and accidental collisions may occur. From these reasons, an unmanned construction system is required and has been desired and eagerly studied. Yamamoto et al. [] roosed an automatic control and motion lanning system using D ground shae ma information for a t-class excavator. This system achieved the automatic target trajectory tracking. Kang et al. [] automatically controlled the D trajectory of a t-class excavator with hydraulic cylinders. Ha et al. [5] demonstrated the trajectory tracking by imedance control for a small (mass:.5t, width: mm, deth: 95mm, height: mm) excavator. Although these systems realized good erformance for the trajectory tracking, unfortunately they cannot be used with a PM because of the following constraint conditions. () A large excavator cannot be used, because the equiment must ass through the material shaft with inner diameter of.8m to be assembled and disassembled in the narrow sealed working chamber. () Machines that use combustion engines cannot be used in the sealed and ressurized working chamber. These constraints hamer the realization of an unmanned construction system for PM. From this oint of view, we develo an automatic excavation system using a caisson shovel oerating in high air ressure and narrow underground sace. Here, we first imlement the sensors into the caisson shovel with five DOFs to measure three rotary joint angles and two rismatic joint lengths and then reroduce the trajectory of soil mountain removal by a skilled human oerator using a roortional-derivative (PD) control for hydraulic cylinders. Finally, we assess the trajectory oyright 7, The Organizing ommittee of the IDES 7
tracking erformance and the motion characteristics. The remainder of this aer is organized as follows. In Section, we describe the architecture of the caisson shovel and the imlemented sensors. In Section, we conduct the kinematics of the link maniulator, mathematically. In Section, we demonstrate the tracking trajectory by a skilled human oerator and assess the erformance and motion characteristics. In Section 5, we conclude the aer and outline future works. Driving body & Rotating frame ounter weight Boom & Arm 77mm Bucket 77mm Man shaft Remove soil Material shaft Teleoeration room Fig. aisson shovel overview arriage traveling Table Movable range mm d 7mm Boom yawing 8 deg 8 deg Above ground Underground aisson box Sinking Boom itching Boom exansion Bucket itching 7 deg deg mm d 9mm 97 deg deg eiling Rail track aisson edge aisson shovel Fig. Pneumatic caisson facilities Working chamber. m Earth bucket aisson shovel. Joint configuration Figure shows a caisson shovel in an unressurized test working chamber. The caisson shovel is disassembled into four units: carriage, boom, counter weight, and bucket units. These units are carried in the working chamber through the material shaft with inner diameter of.8m, assembled, and after construction comletion, carried out again through the material shaft. The caisson shovel is a heavy machine with a weight of t and is modeled as a link maniulator with characteristic five DOFs: carriage traveling on the rail track (d ), yaw rotation of the boom unit (θ ), itch rotation (duming) of the boom unit (θ ), exansion and contraction of the boom unit (d ), and itch rotation of the bucket unit (θ ), as described in Section. Table lists the movable range for each DOF. Although a common backhoe [6] is driven using hydraulic cylinders by a combustion engine and consists only of rotary joints, the caisson shovel is driven using hydraulic cylinders through an external electric ower suly and has two rismatic joints and three rotary joints. Moreover, while the former moves on the ground surface and suorts the reaction force using the ground, the latter travels on a rail track and suorts the reaction force using the rigid rail track on the ceiling. In addition, the carriage unit is fixed on the rail track by a locking mechanism during excavation (d is constant).. Sensor imlementation Figure shows sensors imlemented to measure the joint variables. The kinematics, e.g., the relationshi between the joint variables and the osition-orientation of the caisson shovel, is uniquely determined. A laser rangefinder (A: DL5-N), rotary encoder (B: TRD-JRZ), laser range finder (: LR-TB5), laser range finder (D: LR-TB5), and a air (forward and backward) of ultrasonic flowmeters (E: FD-Q) are used to measure the osition of the carriage on the rail track (d ), the boom yaw angle (θ ), the boom itch angle (θ ) from the dislacement of the cylinder, the boom exansion and contraction (d ), and the bucket itch rotation (θ ), resectively. Since an individual ultrasonic flowmeter is used to measure the unidirectional flow quantity, the bucket itch angle is calculated by integrating the sensor value and multilying the ie diameter of the hydraulic cylinder. In this study, we chose such an indirect sensor caable of being mounted on the boom unit far from the bucket unit. Because, although the rotary encoder is useful and accurate, the bucket, i.e., the joint axis onto which the encoder is mounted, is sometimes submerged under muddy water. A: Laser range finder (d ) E: Ultrasonic flow meters(θ ) : Laser range finder(θ ) Fig. Imlemented sensors B: Rotary encoder(θ ) D: Laser range finder (d )
.. Investigation of bucket itch rotation accuracy The ilot exeriments revealed that the accuracies for the joint variables, d, θ, θ, and d, were sufficient for excavation oerations, because these errors were less than mm, deg, deg, and mm, resectively. However, since the bucket itch angle is indirectly calculated based on the forward and backward flow quantities through the hydraulic cylinder, we investigated the accuracy exerimentally, as follows. We first rotated the bucket from the lowest angle (-97deg) to the highest angle (eg), then rotated inversely to the lowest angle, and reeated this rocedure five times. Figure shows the time history of the forward (blue) and backward (red) flow quantities integrated every.s (right axis) and the bucket itch angle (urle) calculated from the sensor values. The light blue and yellow green ines are the geometric uer and lower limit angles corresonding to above lowest and highest angles, resectively. The average flow quantities at the end of the ascending (deg) and descending (-97deg) motions were.7l (standard deviation: std..7l) and.8l (std..6l), resectively. The calculated bucket itch angles were 5.9deg (std..deg) and -86.9deg (std..7deg), resectively. Based on these results, we found that since the standard deviation was low, the reroducibility was high. However, since the bucket itch angle was overestimated during the ascending motion and underestimated during the descending motion, hysteresis existed. This is assumed to be due to the effect of gravity. The ascending motion requires a larger torque than the descending motion to counter gravity. larification of this mechanism is a subject for future investigation. Here, we comensated these values by multilying the ascending and descending coefficients, i.e., roortionality constants:. and.. Figure 5 shows the time history of the comensated forward and backward flow quantities integrated every.s and the bucket itch angle calculated from the comensated sensor values. Naturally, the comensated values at the end of the ascending and descending motions corresonded with the theoretical values of deg and -97deg. The intermediate value will need to be investigated. Bucket itch angle [deg] 6.... -. -. -6. Forward instant flow quantity Backward instant flow quantity Bucket angle -8. Lower limit: -97 deg -.. 5.. 5.. Time [s] Fig. Time history of flow quantity and calculated bucket itch angle.8.6 Uer limit: deg... -. -. -.6 -.8 Flow quantity [L] Bucket itch angle [deg] Fig. 5 Time history of comensated flow quantity and calculated bucket itch angle aisson shovel model. Mathematical model and kinematics Figure 6 shows the mathematical link model for the caisson shovel and the coordinate systems. This is the initial configuration when d d. Here, R is the frame of reference set to the working chamber ceiling and E is the coordinate system for the bucket claw. Table shows the link arameters [7] from to E. Here, ai-, αi-, di, and θi mean the translation and rotation around X axis and the translation and rotation around Z axis, resectively. The homogeneous transformation matrix is obtained from the link arameters and the relationshi between R and, as follows: S TE S where R x y z 6.... -. -. -6. -8. Lower limit: -97 deg -.. 5.. 5.. Time [s] 6 ( l S S ( l S l 6 6 S l 5 l 5 l S 5 S S S l S ( d x y z l ( d l ( d l ) S l ) S l ) () l ) d l l ) Here, Si sini and i cosi. From this, the relationshi between the osition and orientation of the shovel claw (bucket ti osition), r, is x E x y E y z E z r Forward instant flow quantity(fix) Backward instant flow quantity(fix).8 Bucket angle(fix).6 Uer limit: deg. -. -. -.6 -.8 () where,, and are the yaw, itch, and roll angles, resectively. Hence, this mechanism cannot change the roll angle. The osition and orientation excet for the role angle are uniquely determined by five DOF variables. The link lengths are as follows: l = 89, l = 7, l = 5, l = 6, l 5 = 8, and l 6 = 59... omensated flow quantity [L]
Rail track l y x d l x z R, x R, z x l arriage unit Boom unit this system is sufficient for the simle soil mountain removal task, other comlex oerations, such as excavation lanned by multile caisson shovels, require otimized satiotemoral control to avoid collisions. The characteristics that individual hydraulic oututs affect mutually and their torques are changed must be also considered. The PD control considering not only the nonlinearity of the caisson shovel with a large inertia and friction but also these issues is very imortant. z l th joint osition y x l 6 l 5 Bucket ti osition z E x E Bucket unit Fig. 6 Mathematical link model for caisson shovel Bucket ti osition th joint osition Table Link arameters i ai- αi- di θi -l θ - dl 9 θ - l 9 d+l -9 θ E l5-9 - l6 Trajectory tracking erformance We investigate the trajectory tracking erformance by reroducing the trajectory during soil mountain removal by a skilled human oerator. Here, the caisson shovel reroduces the trajectory using three DOFs: boom itching (θ ), boom exansion and contraction (d ), and bucket itch rotation (θ ) on the sagittal (x-z) lane by the PD control. The samling time for the feedback control is s. The outut of the hydraulic cylinder, i.e., the oenness of the roortional solenoid valve, was set to 9% for the boom itching, exansion, and contraction and to 8% for the bucket itching based on ilot exeriments. Figure 7 shows the excavation trajectories for the bucket axis and claw by a skilled human oerator using the PD control. The horizontal line at -mm indicates the distance from the ceiling of the working chamber to the ground surface. The trajectory tracking by the PD control overshot the target osition when the caisson shovel started to move, because the inertia of the caisson shovel was very large. Moreover, the error for the bucket claw was larger than that for the bucket axis because of the error in the ultrasonic flow sensor described in the revious section. Since the static friction for the heavy machine was very large and differs with the dynamic friction, the control value for the target became nonlinear and comlex. In this case, the boom unit started exanding from the outut of 5% and contracting from the outut of 5%. That is, the force by the outut of 5% means the maximum static friction. Figure 8 shows a hotograh of soil mountain removal by the PD control. Although the accuracy of Fig. 7 Excavation trajectories for a human oerator and PD controller 5 onclusions In this study, we develoed an automatic excavation system using a caisson shovel oerating in high air ressure and narrow underground sace. Here, we imlemented sensors to a caisson shovel with five DOFs and reroduced the trajectory for a soil mountain removal task by a skilled human oerator using a PD control for hydraulic cylinders. The exerimental results indicated that the ultrasonic flowmeter had hysteresis caused by gravity. This was comensated by the roortionality constants for the ascending and descending. Although the accuracy of our system was sufficient for simle oeration, trajectory tracking by the PD control tended to overshoot, because the inertia and friction of the caisson shovel are very large. The PD control considering the nonlinearity and the characteristics of the caisson shovel is a subject for future work. Moreover, we intend to analyze the kinematic characteristics based on the maniulability and maniulating force and to erform cooerative excavation using multile caisson shovels.
Acknowledgement This research was suorted by a grant through the Sentan Sangyo Souzou Project of Saitama Prefecture and by Sankeikogyo or. The authors would like to exress their dee gratitude to all involved in the research roject. References [] htt://www.orsc.co.j/english/tec/newm_v/ncon. html. [] Kodaki, K., Nakano, M., and Maeda, S., Develoment of the automatic system for neumatic caisson, Automation in onstruction, Vol. 6, (997),. -55. [] Yamamoto, H., Moteki, M., Hui, S., Ootuki, K., Yanagisawa, Y., Sakaida, Y., Nozue, A., Yamaguchi, T., and Shin ichi, Y., Develoment of the Autonomous Hydraulic Excavator Prototye Using -D Information for Motion Planning and ontrol, IEEE/SIE International Symosium on System Integration, (),. 9-5. [] Kang, S., Park, J., Kim, S., Lee, B., Kim, Y., Kim, P., and Kim, H., Path Tracking for Hydraulic Excavator Utilizing Proortional-derivative and Linear Quadratic ontrol, IEEE onf. on ontrol Alication, (),. 88-8. [5] Q.P. Ha., Q.H. Nguyen., D.. Rye., H.F. Durrant-Whyte., Imedance control of a hydraulically actuated robotic excavator, Automation in onstruction, Vol. 9, (),. -5. [6] Yamamoto, H., Uesaka, K., Ishimatsu, Y., Yamaguchi, T., Aritomi, K., and Tanaka, Y., Introduction to the General Technology Develoment Project: Research and Develoment of Advanced Execution Technology by Remote ontrol Robot and Information Technology, ISAR, 6. [7] J. Denavit and R.S. Hartenberg, "A Kinematic Notation for Lower-Pair Mechanisms Based on Matrices", ASME Journal of Alied Mechanics, Vol. 77, (955),. 5-. Received on December, 6 Acceted on March 9, 7 Fig. 8 Soil mountain removal by the PD control