Multiple sensor fault-tolerant predictive control for autonomous surface vehicle formation

To address the problem of cooperative Autonomous Surface Vehicle (ASV) formation control under multiple sensor faults (e.g., abnormal positioning and speed data), this paper proposes a predictive fault-tolerant control approach based on the Distributed Extended Kalman Filter (DEKF) and Model Predictive Control (MPC), termed DEKF-MPC. An ASV formation path following model is established based on parameterized paths and virtual leader-follower structure, and the ASVs' cooperative objective for formation is transformed into a specific trajectory tracking task for each ASV. Moreover, a DEKF state estimator based on Statistical Analysis with Random Sample Consensus EKF (SAR-EKF) and Short Term Memory EKF (STM-EKF) is designed by integrating multiple sensor faults into the DEKF process model and taking auxiliary data from fault-free ASVs as inputs to DEKF, to estimate the ASVs' position, Speed Through Water (STW), and current speed. Then, the DEKF-MPC formation path following fault-tolerant controller is designed with a defined cooperative control objective function, estimated states, and multiple constraints. The simulation results show that the proposed DEKF-MPC approach outperforms the Adaptive Particle Filter-MPC (APF-MPC) and Random Sample Consensus-EKF-MPC (RANSAC-EKF-MPC) methods. It enables ASVs to accurately track target trajectories and maintain consistent headings, especially in the presence of multiple sensor faults.