Motion Planning of Autonomous Vehicle for Illegal Parked Vehicles at No Parking Area

It is an inevitable challenge that autonomous vehicles face the scenes of on-street illegal parking. In this paper, we propose a motion planner of both path and speed based on machine learning methods to imitate behavior of driver driving past illegal on-street parking. On-road free space and vehicle dynamics are detected for learning features to generate suitable trajectories. The trajectory generation is further integrated with a state-space decision making system. Our proposed method is validated on a real autonomous vehicle in the presence of illegal on-street parking with an on-road test. Result shows that the planning framework gives a safe trajectory to avoid objects for considered illegal on-street parking with a guarantee that a collision will not occur.