展館導(dǎo)游機(jī)器人輪式行走系統(tǒng)設(shè)計,展館導(dǎo)游機(jī)器人輪式行走系統(tǒng)設(shè)計,展館,導(dǎo)游,機(jī)器人,輪式,行走,系統(tǒng),設(shè)計 RocHdllqr Of lh Seallie. Washington June 1995 TP14 - 450 Efficient Control of a Mobile-Task Robotic System J.M. Lee, M.C. Lee2, K. Son2,D.S. Ah3, S.H. Han4, and M.H. Lee2 1. .Department of Electronics Engineering 2: Department of Mechanical Enginering Pusan National University Pusan, 609-735, Korea Fax: (8251) 515-5190 jmlee hyowon.pusan.ac .kr Pusan National University :Pusan Fishery University 4:KyungNam University 3 Abstract The control of a mobile-task robot is implemented with the suitable decoupling of the dynamic interacti- ons between two local robots, i.e., a task robot and a mobile robot. The mobile-task robot is a redundant robot formed by the serial connection of the two ro- bots, and it is suitable for the multiple task execution without any constraints on the workspace. The decoup- ling process can be initiated when the motions and static forces required for each robot are described explicitly. In this paper, we demonstrate the perform- ance of a redundant robot system controlled by two independent controllers which are coorperating to achieve a common task. By using the two independ- ent controllers, we can achieve a task assigned at the end of the task robot with high speed and high precision. One typical example of task distribution is that fast and rough motions are mainly assigned to the mobile robot, while slow and accurate motions to the task robot. Forming a redundant manipulator thr- ough the serial connection of two robots, decoupling the dynamics of the two robots through the task distribution and controlling each of the local robot by an independent controller is a new approach for the design/control of high performance robotic systems. 1. Introduction A conventional manipulator has 6 joints for 6 d.0.f motion with the fixed base. With this 6 d.0.f mani- pulator, the end-effector assumed to be located at any position/orientation within its own workspace. However, practically there are some geometrical con- straints, Le., singularity avoidance and/or obstacle avoidance, which may cause the manipulator not to be able to perform a given task or may cause the task execution performance to be very low. Espec- ially, when the trajectory for a given task goes through the singular configuration, the positional con- trol error becomes large and high joint torques are ipulators are introduced by many of researchers 3,41. In designing redundant manipulators, a lot of factors need to be added to the geometrical constrai- nts : 1. The manipulator is contacting with the unknown environment during the guarded motion. 2. The manipulator is required to perform various tasks. 3.The manipulator should maintain high task execution performance in the sense of accuracy and speed. 4. The manipulator should be easy to be handled and maintenanced by the users. These kinds of redundant manipulators are implemented by the various types of manipulators, i.e., by adding a joint (either a prismatic or a revolute joint) to the conve- ntional robots or by forming the parallel manipulator stmcture U, for example, Stewart Platform. Recently instead of the fixed base robots, the moving base robots are attracting high interests, since these kind of robob have a lot of advantages compared to the fixed base robots. First of all, the moving base robots have unlimited work space as far as the base can be moved arbitrarily. And also, by the movement of the base, the singularity of the robot can be avoided. However, in the behind of these advantages, we have additional job that is to control the base of the robot, i.e., to control a mobile robot which supports the base of the robot When we use this kind of moving base robot, the upper robot which performs a task tm be termed as ”Task Robot” and the lower robot which moves the base of the task robot can be termed as “Mobile Robot”. The coordination between 1,hese two local robots is a critical point of the control of this mobile-task robot system. There are lots of types of mobile-task robot systems conceptually 11,611. Forming the redundant robot by the serial connection of the two independe- nt robot is a new trial for the redundant robot design. This saves much cost in designing redundant robots. Also, we can change the task rob- ot for a specific type of task easily. We are going to show the design procedure for the mobile-task robot and finally to demonstrate the performance of the mobile-task robot system. 2. Kinematicsof a Mobile-Task Robot required, which cause the controi to be unstable. To overcome these kinds of constraints, redundant man- 2805 To achieve a unified control of the mobile-task robot for a common task, the whole system should ?=- Pic. 1. Mobile-Task Robot Systemby Serial Conne- ction of Two Independent Robots be analyzed and controlled in a common frame, since we serially connected a mobile robot and a task robot which have their own controllers individually. Picture 1 shows the mobile-task robot system deve- loped by this research. Fig. 1 shows the link frames assigned onto the mobile-task robot system and a world frame assign- ed onto the ground. U following constraint equation. Yt cos P - xtsinp = 0 (1) it cos qJ + Ytsinu, - b p, = I e (3) xr cosp + ytsinp, + b 6 = I e, (2) where r is the radius of wheels. We assigned a frame (1) at the end-plate of the mobile robot, which will be used as a base for the task robot. The positioning vector from the world frame to the frame (11, e=Px Py PzIT, will be rep- resented as functions of joint variables, 81, (10) 0 00 where fT n: I is the 6 dimensional force at the first joint of the task robot, TO is the joint torque required at the last joint of the mobile robot. Therefore, even though we combined the two local robots and formed a redundant robot, we can implement high speed control of the redundant robot, as far as two controllers are used with the dynamic decoupling between two robots as in our scheme. 4. Composition of a Mobile-Task Robot The control system has a hierarchical architecture ; 1. A main controller governs both of the mobile and task robots for the higher level processes, for example, trajectory planning, obstacle aviodance and cooperation (task distribution) between two robots. 2. Each joint controller governs the joint control. The control architecture of mobile-task robot system is shown in Fig. 2. The main controller (EM-PC) calculates the joint angles - using the inverse kinematic equations - for each joint of the mobile-task robot for the desired positiodorientation of the end-effector, and sends the values to the joint controllers. Note that the end-plate of the mobile robot is the same as the base of the task robot. The joint limits and task execution perf- ormance of the task robot are monitored by the main controller to access the coordination of the mobile robot if necessary. The main controller also performs the trajectory planning of a given task considering the capability of joint controllers as explained in the previous section. The variational values of the joint angles are finally sent to the joint controllers for the instantaneous velocity command. Since the task robot is activated by 5 stepping motors with a control unit., the main controller sends I -I Tash robo: -7 I: 7 Fig. 2. Architecture of Mobile-Task Robot Control System to the task robot the step numbers for each joints for each task point through the bi-directional data bus. The mobile robot is composed of three DC servo motor controllers for the x, y and z directional linear motions. Each motor is controlled to track the velocity value - sent from the main controller throu- gh the bi-directional data bus by every T, msec - according to the PD algorithm. The joint angle values read by the encorders are sent back to the main controller to monitor the control performance of the mobile robot and also to modify the task trajectory based upon the tracking errors. Therefore, the main controller forms a control loop for the whole mobile robot movement. The mobile robot system is composed of four major parts: a control box unit, the mechanical robot, the interface unit, and the main controller. In Fig. 3, the composition of the mobile robot is represented as a block diagram. The driving mechanism is composed of two parts:l. Two DC servo motors which drive left and right Fig. 3. Block Diagram of Mobile Robotic System wheels individually to obtain the motion on the x-y plane. 2. One motor which performs the z directional linear motion through a screw gear. The detailed mechanical analysis of the mobile robot is given in the reference lo. Each of the three joint controllers are implemented on each single board and plugged into a mother board designed for the connections of the joint controllers to the IBM PC bus with high expandability 6-91. As it is shown in Fig.3, a motor driver is provided for each motor and an incremental encoder is used for sensing the joint values of each motor. The joint controllers share the bus interface card for the data communication in between the main Controller and 2807 joint controllers. In addition to these parts, power units for motor drivers and digital circuitry are included in the mobile robot control system 131. 5. Preliminary Experimental Results To demonstrate the task distribution between the two robots, we selected a task drawing a big recta- ngular shape which can not be done by only the task robot. To draw a line from the left end to the right end (It is out of the work space of the task robot, unless the base is moved.) on the wall, the base of the task robot needs to be moved to a suitable position. For this purpose, the mobile robot carries the task robot which is attached on the end-plate of the mobile robot to a suitable position. In detail, we used the condition number (which is defined as the ratio of the largest eigenvalue to the smallest eigenvalue of Jacobian matrix, J of the task robot) as a cost function to guide the task distribut- ion between the task robot and the mobile robot. To make the problem simple, we operated the mob- ile-task robot within relatively slow speed range so that the dynamic effects can be ignored in the cont- rol. The experimental procedure can be summarized as follows: 1) Obtain the values of 81 82 83 84 85 1 from the desired positiodorientation of the end-effector, 2) For the obtained configuration, the manipulability ellipsoid 2,3 can be determined for the x-z planar motion. From the ellipsoid, we obtain the value of condition number (C.N.). If this value is smaller than the reference value (it is determined heuristically), the task robot performs the task, i.e., moving the end-effector to the positiodorientation. 3) If the C.N. is larger than the reference value, the robot configuration assumed to be bad to perform the task. Therefore, the mobile robot moves the base of the task robot. At this moment, to keep the end-effector of the task robot at the same position / orientation (stationary state), the motion of task robot should compensate for the motion of its base generated by the mobile robot. This coordinative control mode should be closely controlled. Px Py Pz a B 7 1. 0 Experimental Operation The mobile-task robot performs a task to draw a rectangular shape formed by the path (Px, Py, Pz) = -+(0,44,43) with the same orientation variables, B=O and =90 , while a gripper is holding a pen rigi- dly. (0,44,43)-+(30,44,43)-+(30,44,0)-+ (-30,44,0)-+(-30,44,&3) 0 Simulation For the given reference value, C.N. = 1.37, we checked the configuration of the task robot and modified the configuration by the coordination of the mobile robot during the task execution along the rectangular path. As it is shown in Fig. 5, the C.N. can be kept below the reference value, 1.37 for the whole work path with the cooperation between two local robots. 1.50 C.N* = 1.37 1 ti 1.20 1.10 Fig. 5. Variation of C.N. on the Whole Work Path at C.N. = 1.37 Experimental Results The task described in the simulation sub-section is performed by the mobile-task robot to evaluate the control accuracy of the mobile robot and to demons- trate the coordination scheme of the mobile-task ro- bot. (1) Experimental Condition The center position of the end-plate of the task robot: (O,O,678)(mm) with reference to the world frame. Task path of the end-effector of the task robot : A rectangular shape formed by the four poin- ts,(Y,Z)=(-30,0,43),(30,0,43),(3O,O,O)and-30,0,0. *Initid posture of tool: Q = 9O,B= 0, 7 = 90 (Euler angle) * Distance between path points : 1.25 (cm) The passive compliance is added for normal direct- ion to the drawing surface, Y-axis, to prevent the pen from tilting out from the gripper, and to protect the pen. In this experiment, the performance of the task robot is checked with the motions on the X-Z plane. (2) Results We checked the cooperation points where the coop- eration between the mobile and task robots are required by the simulation and represented in Fig. 6 as numbers from 1 to 12. During the experiment, the Coordinative control is required and performed at the same points (point 1 to 12). Fig. 7 and Fig. 8 show that the positional control errors of x direction, and norm of errors for each of 2 A Fq. 6. Cooperation-occurring Points of Two Robots on the X-Z plane the mobile, the task and the mobile-task robot at the cooperation points ( 1 to 121, individually. 2808 5.00 A Task 0 MOBlb Y x MoblIe-Task 0 -5.00 - 0.00 E, v Fg. 7. Tracking Errors on the Cooperation Points of Mobile-Task Robot in the X-axis Direction (3) Discussion We demonstrated that the coordination between the two local robots can be performed to maintain the configuration of the task robot 0 MoblIe x MObiIe-Task 1.00 - 0 12 3 4 5 8 7 8 9 1011121J Coop. points Fig.8. Tracking Errors on the Cooperation Points of Mobile-Task Robot to be efficient to execute the task. The C.N. is used as a cost function to guide the cooperation between the two robots. By the experiment, we recognized the positional tracking error of the mobile robot is kept within a few mm through the whole task. This accuracy is not good enough for the commercial robot. However, as it can be noticed from Fig. 7-8, the position error of the mobile-task robot is mostly caiised by the task robot which is designed for commercial purposes. We would like to note one of the interesting point obtained by the experiment: The positional control accuracy of the task robot is decreased with the inc,rease of the C.N. Therefore, we can conclude that the C.N. value is well suited for our demonstration. To improve the control accuracy of the mobile-task robot, we do need to develop a control algorithm which suits with the dynamics of the mobile robot and compensates for the disturbances from the wheels. The task robot needs to be replaced by a high accurate robot to achieve high speedhigh accuracy motions as intended. 6. Conclusion In this paper, we demonstrated a way of forming reclundant manipulators by the simple serial connect- ion of two robots: a mobile robot which can provide 3 dimensional linear motion through two wheels and one screw-gear, and a task robot which can be selected and attached at the end-plate of the mobile robot for a specific purpose. For this, we designed and implemented a mobile robot system and tested the performance of the mobile robot in the mobile- task robot system. For the mobile robot, 3 individual joint controllers are designed and implemented. Each controller is governed by the 16 bits micro-controller, Intel N8097BH. To perform the task more efficiently, we selected the manipulability as a cost function, and optimized the configuration of the task robot during the task execution. The results weire satisfactory even though a couple of improvements are required. For the con- trol of the mobile robot, we implemented the PD control algorithm which was suitable to overcome the small distortions, for example, the friction of mechanical joints and the friction of ball-caster with the surface. However, if the task robot is changed during the control process, the distortion might be too large for the control system to compensate for. We do need to develop more robust control algorithm to make the mobile robot suitable for the base of various task robots. 7. References 11 H. Seraji, ”Motion Conlsol of Mobile Mainipulato- s,” Proc. IEEE Conf. on Intelligent Robots and Syst- ems, pp.2056-2063, 1993. 121 T. Yoshikawa, “Manipulability of Robotic Mechanisms,” Proc. 2nd Int. Symp. on Robotics Research, Kyoto, Japan, 1984. 31 SLee and J.M.Lee, ”Multiple Task Point Control of a Redundant Manipulator,” IEEE Conf. on Robotics and Automation, May 1990. 41 K.K.Degaldo, M.Long, and H.Seraji, ”Kine- matic Analysis of 7-DOF Manipulators,” Int. Joml of Robotics Research, Vol.11, No.5, 1992. 51 J.J. Craig, Introduction to Robotics - Mechanics and Control, 2nd edition, Addison-Wesley Publishing Company, New York, 1986. 61 M.W.Spong and M.Vidyasagar, Robot Dynamics and Control, John Wiley & Sons, New York, 1990. 61 D.G.Bihn, ”Universal Six- Joint Robot Controller,” Control System Magazine, pp.31- 35, 1988. 71 Intel Corporation, 16-Hit Embedded Controllers, Embedded Applications - microprocessor and Periph- eral Handbook, V01.2, O.N. 270250-005,1990. SI Intel Corporation, C-96 Users Guide for Dos System, 0.N.167632-002, 1987. 91 Intel Corporation, MCS-96 Macro Assembler Users Guide, O.N. 122048-001, 1983. lo JinHwa Ju and Jang M. Lee, “Implementation of a Redundant Robot Through the Serial Connection of a Mobile Robot and a task Robot,”, hoc. of The First Asian Control Conference, 1994. E111 J-P.Merlet, ”Direct Kinematics and Assembly Modes of Parallel Manipulators,” Int. Journal of Robotics Research, Vol.11, No.2, 1992. El Yoshio Yamamoto and Xiaoping Yun, “Modeling and Compensation of the Dynamic Interaction of a Mobile Manipulator,” IEEE Coference on Robotics and Automation, pp.2187-2192. 1994. 131 CSG LEE and P.R.Chang, “Efficient Parallel Algorithm for Robot Inverse Dynamics Computation,” IEEE Transaction on Systems, Man and cybernetics, pp.532-542, 1986. 2809