Rao-Blackwellized Particle Filter with Grid-Mapping for AUV SLAM Using Forward-Looking Sonar

Abstract

This paper adresses underwater localization for an AUV using SLAM and Forward Looking Sonar (FLS) data. The proposed method is Rao-Blackwellized Particle Filter (RBPF) with grid mapping. The original RBPF uses multiple suggested trajectories, where each of which has a occupancy grid map. Analogously, range measurements from the FLS are computed for each trajectory. A scan matching procedure is then performed for the measurements of each trajectories and the best matches are then chosen. The performance is confirmed through simulations and experiments. The suggested method enables SLAM in enclosed underwater environments using noisy FLS measurements without any type of artificial landmarks or assumption on the type of environment, and without any increment in runtime complexity for long operation times.