An efficient indoor positioning particle filter using a floor-plan based proposal distribution

Abstract

We present a novel floor-plan and PDR (pedestrian dead reckoning) based proposal distribution for indoor positioning particle filtering. Including floor-plan information in the proposal distribution makes the particle filtering more efficient than using the map only in the measurement model, because the proposal distribution becomes more accurate and the measurement model less accurate. The method uses offline- computed distances from each point of a regular grid to the closest wall in each direction. Our simulations show that the novel proposal distribution combined with a floor-plan and PDR based motion model improves the positioning accuracy with small numbers of particles and noisy PDR compared to the particle filters that use the floor-plan only for particle weighting.