The Effect of Maps-Enhanced Novel Movement Models on Pedestrian Navigation Performance

In this paper the question of “how much the a priori knowledge of floor plans will assist navigation in challenging indoor environments?” will be investigated. Additionally, the effect of different levels of accuracy of floor plans on the overall estimation process will be studied. In particular, the effect of wrong a priori knowledge of maps and floor plans on navigation accuracy will also be illustrated. The knowledge of floor plans is used to make the prediction step of sequential Bayesian estimation [AMGC02] more realistic. A combination of two movement models assisted with the knowledge of floor plans, operating at the microscopic level and suitable for pedestrian navigation are developed to be used in the prediction step. The constituents are a Map-Enhanced Stochastic Behavioral Movement Model to characterize more random motion and a Diffusion Movement Model to characterize a geographic goal a pedestrian might walk towards. A top-level Markov process is used to determine whether to currently use the stochastic behavioral or the diffusion model; therefore, the model switches between motion that is more goal oriented (diffusion model) or stochastic. Both models use the a priori knowledge of floor plans differently. The technique through which the knowledge of floor plans is included in each of the two movement models will be explained. Additionally, the effect of the knowledge of various densities of walls on the overall estimation process will be analyzed.