Human motion prediction for navigation of a mobile robot

Human motion prediction is an important feature to improve the path planning of mobile robots. An exact prediction of the pedestrian trajectory allows not only to avoid them safely but to induce also socially acceptable motions. In this thesis we introduce a procedure to model social rules, which guide pedestrians through crowded human-lived environments. Human motion is always driven by a future destination and is influenced by the distance to other pedestrians and static obstacles. All this parameters are included to estimate the most probable trajectory, selected from a predefined set of human-like trajectories, the so called pedestrian ego-graph (PEG)