Multi‐AUV Cooperative SLAM With Depth and Intensity Dual‐Channel Joint Registration

ABSTRACT The leader‐follower multi‐autonomous underwater vehicle (AUV) fleet equipped with multi‐beam echo sounders (MBES) has become the most effective platform for large‐scale seabed topographic mapping and underwater target localization. Using kilometer‐level high‐speed underwater acoustic communication, compressed terrain data can be shared among multi‐AUV, and bathymetric SLAM (BSLAM) can then yield both precise location and consistent mapping results for the cooperative system. However, BSLAM is prone to failure due to the typically smooth seabed terrain. This paper proposes a depth‐and‐intensity‐fused cooperative SLAM (DI‐COSLAM), which enables long‐term, high‐precision navigation for follower vehicles by fusing inertial navigation system (INS), acoustic ranging, bathymetric depth, and backscatter intensity data. More specifically, the loop closure detection with dual‐channel joint registration (LCD‐DCJR) algorithm is proposed to construct accurate associations, while a multi‐constraint adaptive weight allocation (MCAWA) is designed to adjust the confidence weights of individual sources, including bathymetric, intensity, range and odometer association. Field experiments have shown the proposed method demonstrates improved performance in both single‐AUV and multi‐AUV formation mapping accuracy, as well as in underwater target localization accuracy, compared to the state‐of‐the‐art algorithms.