VorSLAM: A new solution to simultaneous localization and mapping

Abstract

This paper presents a new solution to the problem of simultaneous localization and mapping (SLAM). Traditional extended Kalman filter (EKF) based SLAM (EKF-SLAM) algorithms describe unknown environments with simple geometric elements, such as points for landmarks. This limits the EKF-SLAM to environments suited to such models and tends to discard much potentially useful data. The solution proposed in this paper makes use of all the collected data and gives a more detailed description to the environment, which is a combination of EKF-SLAM and scan match. Landmarks are extracted from raw observations and their locations are estimated by using feature based EKF-SLAM. Around each landmark a local dense map of the environment is built. The landmarks and local maps together give a detailed and compact description of the environment. Voronoi division has been used to build local maps. It guarantees the local maps have none overlaps and have a proper metric scale. Experimental result demonstrates the efficiency of the algorithm.