Indoor localization system based on wearable posture sensors with incomplete observations

Abstract

Radio Frequency Identification (RFID) based indoor localization becomes a hotspot in the robotic research field recently. To overcome the shortcoming that plentiful tags are required in a normal RFID based localization system, this paper presents an indoor localization method by fusing measurements from wearable posture sensors and the absolute position information from scattered RFID tags. From the posture sensors, we can obtain the relative indoor localization data by summing up the vectors composed of step length and heading direction. Since this relative localization is highly affected by the cumulative error, the absolute positions of RFID tags are used as corrections if they are found within a read-range to the user. Because the RFID tags are sparsely placed in the indoor environment, the corrections can be achieved only at incomplete time instants. Therefore, a revised Kalman filter with incomplete observation is applied to the sensor fusion between the posture sensors and RFID tags. Experimental results show that the cumulative error of the system can be significantly reduced and the localization accuracy is enhanced through the sensor fusion.