车路协同条件下智能网联汽车一体化决策模型

Integrated Decision Model for Autonomous Vehicles Based on Cooperative Vehicle Infrastructure System

为了探索当前有限数据条件下面临的无限交通场景问题,提出车路协同条件下基于深度强化学习智能网联汽车决策模型。利用Actor-Critic机制,以highway-env为数据来源,抽取144 h交通数据作为训练数据并进行验证,分析了智能网联汽车在不同车道数条件下的驾驶行为。结果显示,本模型汽车行程时间减少20%以上,碰撞概率减少25%以上,换道轨迹可以通过动力学跟踪。

In order to cope with infinite number of possible traffic scenarios by using limited data,a decisionmaking model for intelligent connected vehicles is proposed based on the cooperative vehicle infrastructure system.By utilizing the Actor-Critic mechanism and the highway-env simulation environment,144 hours of traffic data are used for training and validation.The driving behavior of intelligent connected vehicles was analyzed for different number of lanes.The results show that the vehicle’s travel time obtained using this model is reduced by more than 20%,the collision probability is reduced by more than 25%,and the trajectory can be tracked by lane-changing dynamics.