Figure 11
From: A human motion model based on maps for navigation systems
Average position error of a PF positioning estimator that is based on the playback of real data collected using a foot-mounted IMU, GPS, and a compass. For the black and the red curves, a second cloud far from the building is added to provoke the case of multimodality. The blue curve is generated with one correct cloud and accordingly shows the best performance with an average error of 1.33 m. The black curve results from the case where only walls are used as a replacement for the proper motion model. As time elapses, the wrong cloud dominates and a high average error of 22.84 m is observed. The red curve results from the case where our proper maps-based motion model is used. As time elapses, the correct cloud dominates and an average error of 2.92 m is observed. At 130 s, the wrong cloud vanishes and the red curve shows similar behavior to the blue curve.