Indoor Positioning Using Efficient Map Matching, RSS Measurements, and an Improved Motion Model

Abstract

Unlike outdoor positioning, there is no unique solution to obtain the position of a person inside a building or in Global Navigation Satellite System (GNSS)-denied areas. Typical implementations indoor rely on dead reckoning or beacon-based positioning, but a robust estimation must combine several techniques to overcome their own drawbacks. In this paper, we present an indoor positioning system based on foot-mounted pedestrian dead reckoning (PDR) with an efficient map matching, received signal strength (RSS) measurements, and an improved motion model that includes the estimation of the turn rate bias. The system was implemented using a two-level structure with a low-level PDR filter and a high-level particle filter (PF) to include all the measurements. After studying the effect of the step displacement on the PFs proposed in the literature, we concluded that a new state with the turn rate bias (a nonobservable state in PDR) is needed to correctly estimate the error growth and, in the long term, correct the position and heading estimation. Additionally, the wall crossing detection of map matching was optimized as matrix operations, and a room grouping algorithm was proposed as a way to accelerate the process, achieving real-time execution with more than 100 000 particles in a building with more than 600 wall segments. We also include a basic path-loss model to use RSS measurements that allows a better initialization of the filter, fewer particles, and faster convergence, without the need for an extensive calibration. The inclusion of the map matching algorithm lowers the error level of the RSS-PDR positioning, from 1.9 to 0.75 m, 90% of the time. The system is tested in several trajectories to show the improvement in the estimated positioning, the time to convergence, and the required number of particles