Control of Robot Manipulators with Consideration of Actuator Performance Degradation and Failures

The problems of control of robot manipulators with actuation degradation and detection of actuation faults are addressed in this paper. An adaptive fault tolerant control method is proposed for robot manipulators to maintain the required performance in the presence of actuation degradation, and the same adaptive algorithm is utilized to detect actuation failures. The proposed method can compensate for the effects of not only the parametric model uncertainty such as payload variations, but also the uncertainty in the actuation torque coefficients. The proposed approach integrates functions of parameter estimation, control law, actuator fault tolerance and fault detection. Theoretical analysis and simulation results have confirmed the effectiveness of the proposed method. Index Terms – Robot fault detection, fault tolerant control, adaptive control, and actuator failures. I. INTRDUCTION Most robot fault tolerant control works are focused on the control of underactuated robots, i.e. one or more robot joints have failed, and actuation at those joints has been lost completely [1,10]. The present paper addresses the problems of actuator performance degradation and its on-line monitoring, as well as failure detection. Typical actuator faults are caused by overheating and mechanical component wear, which may not happen as abrupt failures. In this work, we consider an actuator fails if its torque coefficient deviates beyond an acceptable limit. The proposed approach is to design an adaptive control scheme that can maintain the required performance in the presence of actuation degradation, and to detect actuator failures when the actuation performance degrades beyond an acceptable limit. Robust and adaptive control approaches for robot manipulators typically deal with model uncertainties such as payload variations and joint friction [5,6,7,11,12]. In such control approaches, the joint actuators are considered as precise torque sources, and robustness against actuator performance degradation is not accommodated. In case of actuator performance degradation, such controllers cannot take appropriate correction actions and may continue to drive the actuator until it is damaged. The actuator dynamics is incorporated in the overall robot dynamic model in some robust and adaptive control methods [8, 13], and parametric model uncertainty in the actuator dynamics are treated together with other uncertain parameters. Since the actuator dynamics is incorporated in the dynamic model, such control * This work is supported in part by NSERC, Canada, through the grants 227633-00 and 229465-00, and in part by Ryerson Polytechnic University. schemes may be straightforwardly extended to detect actuator faults. However, since the overall dynamic model has higher order, the synthesis of control algorithm is complicated, and the measurement for joint acceleration [8] or the derivative of actuator current [13] is required. As reliability and safety are becoming more and more important in the operation of autonomous and intelligent robotic systems, it is desirable that a robot controller can monitor the actuator performance on-line and detect actuator faults. Depending on the task and environment of the robot manipulator, the robot joint with actuator performance degradation may be kept operating, or be put into a safe mode to avoid unnecessary damages. Fault detection and isolation (FDI) has been extensively studied for industrial process control [4]. Robot fault detection recently attracted significant research interests. In [10], the position and velocity tracking errors are used to detect joint failures. A robot joint is detected failed if a selected combination of the position error and velocity error is larger than a predefined threshold. While it is possible to define such a threshold for a specific trajectory after some online tuning, it is generally difficult to define a globally efficient error threshold because the robot tracking error depends on the control law types and parameters, the payload variation, the position and orientation of the robot manipulator, etc. The cause for larger tracking errors might not be a joint failure. In [2] observers are developed for fault detection. Residuals are generated by comparing measured system outputs and those predicted by the observers. However, in the observer-based methods, the estimation or measurement of joint acceleration is required. Recently, in [3], an actuator fault detection method is formulated using full manipulator dynamics. The torque estimate is filtered in order to eliminate the need for joint acceleration measurement. Adaptive and robust techniques are then applied to deal with model uncertainties. In this paper, an adaptive control approach is proposed to maintain the desired performance in the presence of actuation degradation, to monitor the degradation on-line, and to detect actuator failures. In the proposed approach, the actuator dynamics is still considered fast enough and can be neglected for the positional trajectory tracking control, but the torque coefficients are incorporated in a new parametric dynamic model formulation with the commanded joint torques as inputs instead of the actual joint torques. Based on the proposed model formulation, an adaptive control scheme is synthesized to compensate for parametric model uncertainties including that in the actuator torque coefficients. The embedded adaptation law is used to update the actuator torque coefficients together with the inertial parameters of the robot Control of Robot Manipulators with Consideration of Actuator Performance Degradation and Failures G. Liu Department of Mechanical, Aerospace and Industrial Engineering Ryerson Polytechnic University, Toronto, Ontario, Canada M5B 2K3 [email protected] 0-7803-6475-9/01/$10.00© 2001 IEEE Proceedings of the 2001 IEEE International Conference on Robotics & Automation Seoul, Korea • May 21-26, 2001