Navigation Multisensor Fault Diagnosis Approach for an Unmanned Surface Vessel Adopted Particle-Filter Method

Fault diagnosis (FD) of unmanned surface vessels (USVs) is an important and challenging task for ensuring system safety and reliability. In this study, a robust navigation strategy with FD based on an adaptive particle-filter (PF) method is designed for USVs, with consideration of multiple fault modes involving sensors and propellers. This paper utilizes the switching-mode hidden Markov model to describe an unmanned vessel with probable fault modes. Then, to implement FD and robust navigation, an adaptive PF including self-evaluation algorithm and adaptive weight-adjustment method is run on the fault model. The self-evaluation algorithm is used to extend the posterior distribution, whereas the adaptive weight-adjustment method is used to enrich the particle diversity and process the particles with small weight. The proposed method is tested via sea trial to demonstrate the effectiveness of the fault model under failure of various sensors and propeller conditions. The results indicate that the proposed approach can realize better FD and generate more accurate state estimation, even when faults occur.