Motion planning in dynamic environment using maneuverability maps

Motion planning in an uncertain dynamic environment is a complex task, especially when obstacles move in a non-linear fashion. In this paper obstacles refer to other vehicles, pedestrians, bicycles, riders etc. In such cases, predicting the obstacles’ kinematics requires a forecast of the obstacles’ trajectories. The choice of the forecast time horizon is critical, especially in conflict scenarios where an accident can be avoided only by adjusting the maneuvers of one of the vehicles. In this paper, we present an approach for establishing optimal forecast time based on Maneuverability Maps that determine a vehicle's possible maneuvers. The approach can be used as guidance for human drivers and can also be implemented in the control system of autonomous vehicles. Simulation results indicate that optimizing the forecast time in a conflict scenario can reduce the probability for an accident.