Utilizing WiFi signals for improving SLAM and person localization

Abstract

This paper describes the use of WiFi signals for improving SLAM and person localization problems. Loop closing is the most important step in large-scale mappings and many previous methods rely on image-based feature matching. Such methods are, however, usually costly and tend to sensitive to illumination variations and the robot heading. We therefore propose a new loop closure detection method using WiFi fingerprints. Since WiFi signals are sometimes quite uncertain, we do matching not a pair of poses but a pair of pose sequences for improving the loop closing performance. We then use the WiFi-recorded map to localizing a person with a smartphone. We develop a particle filter-based method and apply it to a robot call system. Experimental results show the effectiveness of the proposed methods.