A revisiting method using a covariance traveling salesman problem algorithm for landmark-based simultaneous localization and mapping

Abstract

This paper presents an efficient revisiting algorithm for landmark-based simultaneous localization and mapping (SLAM). To reduce SLAM uncertainty in terms of a robot’s pose and landmark positions, the method autonomously evaluates valuable landmarks for the data associations in the SLAM algorithm and selects positions to revisit by considering both landmark visibility and sensor measurement uncertainty. The optimal path among the selected positions is obtained by applying the traveling salesman problem (TSP) algorithm. To plan a path that reduces overall uncertainty, the cost matrix associated with the change in covariance between all selected positions of all pairs is applied for the TSP algorithm. From simulations, it is verified that the proposed method efficiently reduces and maintains SLAM uncertainty at the low level compared to the backtracking method.