Human navigational intent modeling

Decision making under human interference is challenging due to human unpredictability. In this project, a probabilistic model to capture human navigational inference is presented. This model includes conceptual final state of human referred as his goal,  confidence ratio as a measure to reflect the credibility of individual hypothesis,  and human deterministic index a factor inspired from discount factor in reinforcement learning. The main contribution of this work is to propose a generic framework to predict the future navigational position of the human given only his current position, heading and velocity. According to reachability analysis, a transferable gird-based value function is pre-computed and used as a reward function to solve the inverse reinforcement learning problem. In the update stage of the algorithm, the tuning parameters get refined based on the latest available information.  Particle filters are employed to represent the goal and human states.