A Novel Near-Optimal Trajectory Planning Method on Consecutive Curved Roads for Intelligent Vehicles

Consecutive curves are a particular type of road scenarios, with the characteristics of dramatical curvature variance and variable road boundary. Trajectory planning of advanced driver assistance systems (ADAS) for continuous curved roads remains challenging. Optimal planning can improve traffic efficiency while ensuring safety, which is worth studying. The paper mainly focuses on optimal trajectory planning under the continuous-curvature scene characteristics. In these scenarios, the road curvature changes significantly, requiring precise consideration of the vehicle’s motion capabilities in trajectory planning. Traditional ADAS trajectory planners always keep the vehicle in the centerline of roads, which fail to accurately constrain road curvature, leading to potential violations of the vehicles kinematic constraints and discomfort. To address these issues, this paper proposes a near-optimal trajectory planning method system, which mainly consists of two modules: 1) By merging the Frenet frame with the vehicle kinematics model, road curvature constraints are accurately applied by us and road boundary constraints are simplified. Utilizing an optimal control strategy, the point-to-point global optimal trajectory planning (PPOGP) for the vehicle is proposed, with its optimality confirmed by comparison with the dynamic programming (DP) as the benchmark. 2) By combining the decision-making method based on DP and the optimization method based on quadratic programming (QP), the vehicles local optimal trajectory planning is proposed. Utilizing the optimal global trajectory as the reference trajectory enhances the decision-making process based on DP, leading to superior obstacle avoidance choices and subsequently improving the vehicles driving efficiency. Finally, the optimality and efficiency of the proposed method compared with the existing ADAS and human-driving strategies are verified by simulations and experiments.