Sparsing of information matrix for practical application of SLAM for autonomous robots

Abstract

As simultaneously localization and mapping (SLAM) is the basis for autonomous robots to realize high level task, it has become a key issue in mobile robotics field to propose a practical SLAM approach for a large scale environment. By analyzing the features of information matrix, this paper presents a novel method to enhance practicability of simultaneous localization and mapping (SLAM) by changing information matrix into a sparse matrix. A large scale environment simulation shows that our sparsification method is highly efficient with the increase of landmarks while maintaining accuracy. Outdoor experiment verifies the promising future of our approach for the application into the real-world.