基于强化学习的车队速度规划与能量管理联合优化

Co-optimization of vehicle queue speed planning and energy management with reinforcement learning

近年来,随着智能交通系统的快速发展,包括车车通信、车路通信等短距离实时无线通信信息,以及道路交通通行信息等远距离交通信息,使得车辆能够实时获知周围车辆运动情况以及前方交通环境情况,有利于提高车辆对周围交通环境的感知能力,以实现合理的出行安排与行驶控制,从而提高车辆的使用性能。为实现车队在多信号灯场景下的节能驾驶,提出了一种基于强化学习的车队速度规划与能量管理联合优化方法。通过SUMO平台,建立了包括5辆汽车的车队通过多个信号灯的场景,结果表明,提出的方法在舒适性、经济性和效率方面均优于传统的驾驶模型(Intelligent Driver Model,IDM)策略。

In recent years,with the rapid development of intelligent transportation system,including vehicle-vehicle communication,vehicle-road communication and other short-range real-time wireless communication information,as well as road traffic information and other long-distance traffic information,so that the vehicle is able to obtain real-time knowledge of the surrounding vehicle movement as well as the traffic environment in front of it,which is conducive to improving the vehicle′s perception of the surrounding traffic environment in order to achieve reasonable travel arrangements and driving control,thus improving the vehicle performance.In order to realize the energy-saving driving of fleet in multi-signal light scenarios,a co-optimization method based on reinforcement learning for vehicle queuq speed planning and energy management was proposed.Through the SUMO platform,a scenario including five vehicles for a fleet of vehicles passing through multiple signals was established.The results show that the proposed method outperforms the traditional driving model Intelligent Driver Model(IDM)in terms of comfort,economy and efficiency.