基于强化学习的智能场外停车诱导

Intelligent outside-area parking guidance based on reinforcement learning

车辆保有量急剧增加导致停车场资源紧缺,针对“停车难”造成城市拥堵日益严重的问题,提出一种优化的场外停车诱导策略。结合车辆与路侧单元通信架构,引入预订机制构造未来停车信息预测模型,在马尔科夫决策过程下应用强化学习方法完成停车诱导策略,并在仿真平台中依据实际需求搭建模拟环境,对几种停车诱导策略进行性能评估。通过比较车辆的平均停车消耗时间、停车花费总路程以及累积停靠数,可知智能场外停车诱导策略能够明显减少停车成本,可对资源进行合理分配。

The rapid increase in vehicle ownership has led to a shortage of parking resources.Aiming at the increasingly serious problem of urban congestion caused by“difficult parking”,an optimized outside-area parking guidance strategy is proposed.Combined with the vehicle-to-roadside-unit communication architecture,the reservation mechanism is introduced to construct the future parking information prediction model,and the reinforcement learning method is applied to complete the parking guidance strategy in the Markov decision process.In the simulation platform,a simulation environment is built according to actual needs,and the performance of several typical parking guidance strategies is evaluated.By comparing the average parking time,total parking distance,and cumulative number of parking vehicles,it is concluded that the intelligent outside-area parking guidance strategy can significantly reduce the parking cost and allocate resources rationally.