Collision-Free Cooperative Motion Planning and Decision-Making for Connected and Automated Vehicles at Unsignalized Intersections
提出一种协同运动规划与决策方法，通过学习型迭代优化和蒙特卡洛树搜索算法，让网联自动驾驶车辆在无信号交叉口无碰撞、高效通过，无需排队。
This article proposes a novel cooperative motion planning and decision-making approach for connected and automated vehicles (CAVs) at unsignalized intersections, where a multivehicle collision-free trajectories generating problem is modeled as a constrained optimization problem. A learning-based iterative optimization (LBIO) algorithm is developed to solve the problem iteratively and obtain velocity-optimal trajectories using the historical vehicle states at previous iterations as data sets. To make the trained trajectories adapt to continuous and time-varying traffic flow, an online decision-making algorithm based on Monte Carlo tree search (MCTS) is presented to derive a time-optimal vehicle passing sequence, where a tree structure is built to efficiently express all possible cluster-dividing modes between vehicles. In addition, we propose a trajectory planning algorithm to regulate velocities of vehicles in the cooperative control area surrounding the intersection. The proposed approach is validated on the SUMO under typical intersection scenarios. Results show that our approach enables potentially conflicting vehicles to go through the intersection simultaneously without queuing and significantly improves the overall traffic efficiency.