Weighted Grid Partitioning for Panel-Based Bathymetric SLAM

Abstract

Bathymetric navigation enables the long-term operation of autonomous underwater vehicles by reducing navigation drift errors with no need for GPS position fixes. In the case that a bathymetric map is not available, the simultaneous localization and mapping (SLAM) algorithm is required, but this increases computational complexity and memory requirement. Panel-based bathymetric SLAM could considerably reduce the computational burden. However, it may suffers from incorrect update when the vehicle does not belong to the updated panel. This study proposes a new update method, called weighted grid partitioning, which considers the probability distribution of a vehicle's location, and is more effective in terms of the map accuracy, computational burden, and memory usage compared to standard update methods. The feasibility of the proposed algorithm is verified through simulations.