Behavior Assessment and Trajectory Planning for Autonomous Vehicles on Structural Urban Roads

For autonomous vehicles driving on structural urban roads, we should not only concentrate on passing efficiency but also address the interaction among vehicles, especially the potential risks from the ambient obstacles. In order to find a more secure trajectory, a hierarchical framework combining behavior planner and trajectory planner is proposed in this paper. Probabilistic risk assessment associated with acceleration is used to obtain the optimal behavior with low risk. The spatiotemporal trajectories are generated using quintic polynomial and model predictive method under the guidance of the adopted behavior. The effectiveness of the proposed method is verified in a simulated scenario.