Movement Prediction Hypotheses for Pedestrians and Trajectory Planning for Cooperative Driving Systems

It is a challenge to find a safe trajectory for automated vehicles in urban environments with pedestrians. The prediction of future movements with 100% certainty is impossible, if the intention is unknown. A Gaussian process approach is used to formulate future movement hypotheses of the pedestrian based on historical movements. A mixed integer linear programming (MILP) optimization approach is used for the trajectory planning of the vehicle. The collision probability between the ego-vehicle and pedestrian is used as constraints in the optimization. This approach is useful for cooperative vehicle systems, with historical movement data in a fixed urban environment (e.g., intersection) and the premise that pedestrians follow typical movement data.