Fault Detection for Autonomous Underwater Vehicles Based on Zonotopic Set-Membership Estimation

A novel sensor fault detection framework based on zonotopic set-membership estimation is proposed for autonomous underwater vehicles in the context of unknown but bounded perturbations. First, the zonotopic propagation and intersection properties are utilized to derive the prediction state set and the measurement state set. Then, two methods, namely, projection and polytopic conversion, are provided to examine whether there is an intersection between these two sets. The intersection checking could be used to ascertain the occurrence of sensor faults. To analyze the detection performance of the proposed methods, a minimum detectable fault set is introduced. Finally, pool experiments are conducted to validate the effectiveness of the proposed methods.