Data-driven model-free resilient speed control of an autonomous surface vehicle in the presence of actuator anomalies.

This paper is concerned with the resilient speed control of an autonomous surface vehicle (ASV) in the presence of actuator anomalies. A data-driven model-free resilient speed control method is presented based on available input and output data only with pulse-width-modulation inputs. Specifically, a data-driven neural predictor is designed to learn the unknown system dynamics of the speed control system of the ASV. Then, a resilient speed control law is designed based on the learned dynamics obtained from the neural network predictor, where a cost function is designed for selecting the optimal duty cycle for the motor. The stability of the data-driven neural predictor is analyzed by using input-state stability (ISS) theory. The advantage of the developed data-driven model-free resilient control method is that the optimal speed control performance can be achieved in the presence of actuator anomalies without any modeling process. Simulation results show the learning ability of the data-driven neural predictor and the effectiveness of the proposed data-driven model-free resilient speed control method for the ASV subject to actuator anomalies.