MAPS AND FLOOR PLANS ENHANCED 3D MOVEMENT MODEL FOR PEDESTRIAN NAVIGATION

Pedestrian mobility models are becoming essential in several technologies and techniques. Applications of these models could be found in the areas of infrastructure design, evacuation planning, architecture, robot-human interaction, pervasive computing or navigation and localization. Within the scope of this paper, the purpose of such models is to realistically represent the stochastic nature of pedestrian's movement. Our aspiration is to generate a "movement" or transition model for positioning systems that are based on sequential Bayesian filtering techniques, such as particle-filtering [AMGC02] [GSS93]. However, the developed models can be applied to many of the above application domains. In this paper the three dimensional pedestrian movement model presented in [KKRA09] is extended in order to make use of the valuable prior knowledge of maps of the surrounding environment. The result is a three dimensional mobility model that is capable of representing pedestrian movement in challenging indoor and outdoor localization environments. Examples of such environments are multi-floor buildings, streets, ways, meadows, coppices and forests. Additionally, some quantitative and qualitative analyses of the model and the improvement it brings to the overall positioning performance will be illustrated. The model actually consists of two movement models, operating at the microscopic level and suitable for pedestrian navigation. The constituents are a Three Dimensional Stochastic Behavioral Movement Model (3D-SBMM) to characterize random motion, and a Three Dimensional Diffusion Movement Model (3D-DMM) to characterize geographical goals a pedestrian might walk towards. In order to account for the fact that humans might switch between a goal-directed motion and a stochastic motion, a top-level Markov process is designed to determine when to switch between the 3D-SBMM or the 3D-DMM. Both models use the a priori knowledge of maps and floor plans. The designed model is implemented, tested and evaluated in an already available distributed simulation and demonstration environment for mobility, localization and context applications. The benefit of movement models in the framework of dynamic positioning estimators and a summary of related work will be discussed in section 1. The three dimensional movement model, its constituents, properties and computations will be explained in details in section 2. The question of "Can maps and floor plans replace a proper movement model?" is discussed in section 3. System design and implementation will be illustrated in section 4. Experimental results will be given in section 5. Finally, some conclusions and future work will be given in section 6.