ADP-based active optimal fault-tolerant control for modular manipulator with multiple sensor failure

Abstract

In this paper, we propose a novel active fault-tolerant control (AFTC) method for modular manipulators with multiple sensor failure, which is based on adaptive dynamic programming (ADP). Combining the principle of diffeomorphism, a first-order filter is introduced to convert sensor faults into pseudo-actuator faults. Combine the sensor faults estimated by the adaptive sensor fault observer to establish the Hamilton-Jacobi-Bellman (HJB) equation. Based on the ADP, the online policy iteration (PI) algorithm is applied to solve the HJB equation by establishing a critic neural network. It is proved that the closed-loop modular manipulators system is asymptotic stability according to the Lyapunov theory. Simulation results of 2-DOF modular manipulator systems with two configurations demonstrate the validity of the proposed algorithm finally.