Moving Horizon Multi-Robot SLAM Based on C/GMRES Method

Abstract

The simultaneous localization and mapping (SLAM) is the task of simultaneous estimation of the position and attitude of a mobile robot and the location of landmarks. This paper is concerned with the distributed multi-robot SLAM problem. In the multi-robot system, the robots share their sensor measurements and past estimates with their adjacent robots. We develop a distributed multi-robot SLAM algorithm based on the moving horizon estimation (MHE) technique. In the MHE technique, each robot computes the above estimates by a numerical minimization of a certain cost function. By including the information shared with the adjacent robots into the cost function, the robots can compensate each other to guarantee good estimation accuracy. The proposed algorithm also employs the C/GMRES method for the above optimization to improve the computational efficiency. The effectiveness of the proposed algorithm is verified through numerical experiments.