WiFi 와 자기장 신호를 이용한 대규모 공간에서의 그래프 기반 SLAM

Abstract

WiFi has the characteristic that MAC address is unique, but the inherent noises and attenuation of WiFi signals over the space make accurate localization difficult. On the contrary, it is possible to accurately localize the pose using the magnetic field, but it has a characteristic that false positive matching is likely to occur due to its ambiguity. In this paper, we propose a user localization method using only the sensors embedded on a smartphone. PDR information is generated using the IMU. WiFi matching is performed using the similarity of WiFi signature, and the WiFi matching results are used as global localization. The magnetic matching boundary is determined from the WiFi RSSI information, and magnetic matching is performed only within the matching boundary. The localization results generated from the PDR are used as odometry constraints in the graph structure, and the matching results generated by WiFi and magnetic signals are used as measurement constraints in the graph structure. Through real experiments, it was confirmed that user-level localization is possible by using the proposed method.