SLAM and a novel loop closure detection for autonomous underwater vehicles

SLAM has increasingly become a hot topic for vehicle applications in underwater environments, searching an accurate and effective method for loop closure detection is one of the key issues for SLAM. The benefit of closed loop is the repeated observation of features, which can update the overall vehicle trajectory and environmental map to get more precise localization and mapping. In this paper we explore a novel multi-restraint method of closed-loop detection based on the time limit, data association and the location deviation. The experimental results of simulation and sea trial verify the validity and feasibility of the proposed method for AUV SLAM in large-scale real world environments.