Pedestrian Dead Reckoning with Turn-Based Correction

Pedestrian dead reckoning (PDR) is a popular localization approach. However, PDR is error prone at turning points. During a turn, a small error in heading estimation can cause severe accumulated positioning errors thereafter. In this paper, we propose a turn-based correction approach which aims to reduce the potential accumulated errors caused by inaccurate heading estimations during turns. We consider a crowdsourcing environment in which previous users' data can help in the positioning of current users. For previous and current users moving in reverse directions and turn at the same position, we match the previous user's turning position with the current user's using Kullback-Leibler (KL) divergence based on the received signal strength (RSS) scans collected. Upon a match, the previous user's pre-turn positions can be used to improve the post-turn positions of the current user. Tests are done for different types of paths and the results show that the proposed approach can effectively reduce the PDR errors caused by inaccurate heading estimations at turning points.