Active fault-tolerant hybrid control integrated with reinforcement learning application to cable-driven parallel robots

Abstract

This paper investigates how to maintain control accuracy in cable-driven parallel robots (CDPRs) when faced with actuator faults and lumped uncertainties. An active fault-tolerant hybrid control (AFTHC) scheme integrated with deep reinforcement learning (DRL) is proposed to address the issue. The AFTHC scheme includes a tracking controller, a fixed-time sliding mode observer for fault detection, and a DRL-based fault compensation controller. The fault compensation controller is activated upon detecting an actuator fault to enhance the system stability and recover the control performance. Simulations and experiments are carried out to verify the effectiveness and superiority of the AFTHC scheme. The results indicate that the AFTHC scheme effectively enhances fault tolerance and rapidly recovers control accuracy.