Adaptive output-feedback fault-tolerant control for space manipulator via actor-critic learning

In this paper, an adaptive output-feedback fault-tolerant control methodology is investigated for the trajectory tracking of space manipulator in the presence of actuator failures, unavailability of the joint velocity, inertia uncertainty and unknown external disturbance. First, to reduce the difficulties in controller design caused by the unmeasured joint velocity of the space manipulator, an effective and simple model transformation scheme based on a first-order filter is proposed to transform the original output feedback system with unknown dynamics into a full-state strict feedback system. Second, by virtue of the combination of a smooth saturation function and the mean value theorem, the negative effects of actuator failures can be effectively addressed. Then, drawing support from the backstepping technique and the reinforcement learning (RL) strategy based on an actor-critic neural network (NN) framework, a novel RL based adaptive control (RLAC) scheme is designed, which not only circumvents the ”explosion of terms” typically arising in backstepping technique, but also incorporates feedback mechanism into the space manipulator control system. Moreover, by utilizing the model transformation and the RLAC scheme, the proposed control method is model-free and insensitive to external disturbance. Finally, the effectiveness superiority of the proposed scheme is verified by numerical simulations.