Figure 9
From: A human motion model based on maps for navigation systems
Particle cloud representation of a multimodal position estimation scenario (two particles clouds) using a PF-based estimator at different time instances in the case that walls are used as a replacement for the proper motion model. The scenario is based on the playback of real data, and the second cloud far from the building is added to provoke the case of multi-modality. As time elapses, particles that belong to the correct cloud cross more walls, get low weights, and get discarded because of re-sampling. The wrong cloud's particles do not cross walls, get good weights, and dominate at the end.