Multi-Floor Location Method Based on Crowdsourcing

The widespread deployment of wireless infrastructure makes the fingerprint location method based on WiFi become one of the most universal location methods. However, the time-consuming and labor-intensive fingerprint database construction hinders the development of RSSI(received signal strength indication) fingerprint localization. In this paper, we propose a low-cost and high-efficiency multi-floor fingerprint database construction method based on crowdsourcing aiming at the difficulty of fingerprint construction. Firstly, the indoor floor plan is transformed into indoor semantic map. Secondly, the data of IMU(inertial measurement unit) in the smartphone of crowdsourcing users are collected, and the sensor data are classified into corresponding floors by KF(Kalman filter) fusion algorithm. A segmented trajectory acquisition method is proposed, according to the sensor data, the relative trajectory and RSSI value sequence of the user are acquired. Finally, HMM(hidden Markov model) and TM-Viterbi(track matching Viterbi) algorithm is used to match the trajectory with the main path of indoor semantic map, thus providing the floor label and physical location label for RSSI value sequence. The HMM map matching algorithm of MCSLoc does not need the user’s initial location. The experimental results show that MCSLoc can quickly obtain the absolute initial position of the trajectory, construct multi-floor fingerprint database effectively, and improve the efficiency of multi floor location.