A Sensor Fusion Method for Mobile Robot Navigation

Abstract

An important and challenging research issue associated with mobile robot navigation is simultaneous localization and map building (SLAM), which means that the mobile robot could estimate its poses in the environment without external information, and simultaneously align the local maps. Various kinds of methods such as odometry measurement, landmark matching, laser range image matching and scale invariant feature (SIFT) based algorithm have already been proposed to solve this kind of research problems, but they suffer from inevitable drawbacks. For example, range image matching may suffer from local minimum problem and thus can not get a good location estimation sometimes, and SIFT algorithm can not work if few SIFT intensity features exist in the environment. Furthermore, the map built by laser range image matching is not good for localization when the robot returns to the prior map built by itself, and the map built by SIFT could not be used for obstacle avoidance and next view generation. To solve these problems, in this paper, we propose a sensor fusion method, which makes use of Dempster Shafer algorithm to fuse both of the laser range information and SIFT features information for SLAM. Through a series of experiments, the proposed method is tested and evaluated