A New Reinforcement Learning Fault-Tolerant Tracking Control Method With Application to Baxter Robot
Jun‐Wei Zhu, Zi-Yuan Dong, Zhijun Yang, Xin Wang
Abstract
The fault-tolerant control problem of the flexible multijoint manipulator is a difficult issue due to its strong nonlinearity and coupling. This article proposes a reinforcement learning (RL) based model-free adaptive fault-tolerant control (MFAFTC) algorithm for the multi-joint manipulator. First, a parameter estimation mode switching mechanism is designed based on the two dimensions of the time axis and the sampling period, where an iterative estimation structure is introduced to identify some key parameters online accurately. Meanwhile, the radial basis function neural network is used to identify the spring interference as well as actuator fault, and a compensation fault-tolerant control strategy is proposed. Moreover, the computation complexity is optimized via designing the critic–actor mechanism with an event-trigger parameter selection strategy. Finally, the superiority and effectiveness of the proposed method are verified by the application to the Baxter robot.