Improving Pedestrian Dead Reckoning Using Likely Paths and Backtracking for Mobile Devices

Abstract

Pedestrian Dead Reckoning is a method estimating a persons path from a known starting point based on length and direction of all performed steps. Measuring these parameters, e.g. using inertial sensors, introduces small errors that accumulate quickly into large distance errors. Knowledge of a building's model may reduce these errors as it can be used to keep the estimated position from moving through walls and onto likely paths. Common indoor localization approaches like particle filters track, verify and re-sample several hundred positioning estimates with each user step, resulting in a comparably high computational load. In this paper, we use backtracking to improve an existing localization system tracking a single localization estimate using a foot-mounted inertial sensor and a smartphone. We show how backtracking a single localization estimate can improve the accuracy of indoor positioning systems and discuss restrictions and disadvantages of this approach. Our quantitative results show a reduction of the positioning error by up to 75% and an average endpoint accuracy of 1.91% of the travelled distance with an average computation time of $356.7\mu s$ on a 2014 Motorola G2 smartphone.