Safety-field based motion planning for proactive autonomous driving in dynamic environment

Motion planning in dynamic environment requires proactive response to other road users' intentions. While much progress has been made in intention-awareness methods, challenges remain for the motion planner to consider the uncertainty of intention estimation as well as other risk factors (e.g. the type and motion state of the obstacle). In this paper, we propose a novel motion planner based on safety-field and model predictive control (MPC). Driving-risk assessment is firstly conducted through safety-field based on probabilistic intention-inference with dynamic Bayesian network (DBN), the optimal controller then generates a feasible trajectory within the prediction horizon. Through simulations based on a cut-in situation and an intersection-passing scenario, the effectiveness of this method is demonstrated in enhancing driving safety with considerate motion planning strategy.