A novel trajectory prediction method for AUVs based on the principal axis with RANSAC and an adaptive fuzzy unscented Kalman filter

Abstract

Autonomous underwater vehicles (AUVs) are widely used to explore ocean resources, and trajectory prediction of AUVs is crucial to the active docking method. To address these two issues, this paper proposes a novel and practical trajectory prediction method based on the principal axis with RANSAC (PARANSAC) and an adaptive fuzzy unscented Kalman filter (AFUKF). The PARANSAC is designed to process binocular camera images and point clouds and provides pose estimation of AUVs to the AFUKF, eliminating the need for additional preinstalled markers on AUVs, thus significantly reducing system complexity. The AFUKF, which can adaptively adjust the measurement noise covariance matrix, is designed to estimate the trajectory of AUVs based on the pose estimation of PARANSAC, and the multistep trajectories of AUVs are predicted on the basis of a practical method. Pool experiments are conducted and illustrate the effectiveness of the PARANSAC and AFUKF methods, further verifying the performance of the novel trajectory prediction method.