IJCAS

Due to the complexity of the underwater vehicle itself and its operating environment, it is difficult for fault detection and diagnosis (FDD) unit to obtain accurate fault information in time, which brings difficulties to the implementation of subsequent fault-tolerant control (FTC). Aiming at this question, a novel FTC method is proposed in this paper, which does not rely on FDD units and does not require any prior knowledge of faults. Firstly, a nonlinear continuous time system will be established to describe the underwater vehicle system dynamics with external disturbance, model uncertainties, propeller saturation and unknown propeller faults. To be specific, the nominal controller and fault-tolerant controller are designed respectively. When there is no fault, a nominal controller is used to realize the trajectory tracking control of the underwater vehicle and a quadratic performance index is used to describe and monitor the system performance. Based on the approximation ability of neural networks, a deep deterministic policy gradient (DDPG) algorithm is used to generate the controller compensation signal when the tracking performance degradation caused by propeller fault exceeds the pre-set threshold, so as to realize FTC. Finally, the effectiveness and feasibility of our proposed scheme are verified by simulation.

Keywords: Deep deterministic policy gradient (DDPG), fault-tolerant control (FTC), trajectory tracking control, underwater vehicles.