自动驾驶电动共享车系统自适应集中调度优化  

Simulation optimization of adaptive relocation for shared autonomous electric vehicles

作　　者：侯博宇 刘洪汛 胡路[1,2] HOU Boyu;LIU Hongxun;HU Lu(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu 611756,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Chengdu 611756,China)

机构地区：[1]西南交通大学,交通运输与物流学院,成都611756 [2]综合交通大数据应用技术国家工程实验室,成都611756

出　　处：《交通运输工程与信息学报》2024年第4期25-36,共12页Journal of Transportation Engineering and Information

基　　金：国家自然科学基金项目(72371209,62203367);四川省国际科技创新合作项目(2024YFHZ0266)。

摘　　要：在共享汽车的车辆调度问题中,传统基于库存理论的阈值调度方法要求当区域车辆数高于阈值时立即调度车辆,产生大量调度任务,降低车辆可用性。为此,本文针对自动驾驶电动共享车(SAEV)系统,基于离散事件仿真框架提出一种车辆自适应集中调度优化方法,改进传统的阈值调度方法。具体而言,首先构建SAEV系统的离散事件仿真模型,该模型以站点为中心进行运营区域划分,把用户需求分为跨区需求和非跨区需求,通过车辆自动接送用户实现灵活服务。然后,在仿真模型中,设定固定时间间隔求解一个基于区域阈值的运输问题,确定一次车辆集中调度方案以代替传统阈值触发的单车调度方案。最后,设计了仿真优化框架和BO-SPSA算法有效求解调度阈值以最大化SAEV系统的日利润。该算法通过贝叶斯优化(BO)对同步扰动随机近似(SPSA)算法的参数进行优化,实现更快速和高效的求解。成都市案例表明:(1)BOSPSA相比其他算法能更快速且高效地求解;(2)小规模与大规模运营场景同时验证了本文所提出的自适应集中调度策略相比传统调度策略能服务更多的用户,减少调度车次数,获得更多运营利润;(3)无论运营场景规模大小,当需求较小时,使用快速充电桩并不能有效地提升系统的盈利能力,但随着需求规模的增加,快速充电桩能够更好地增强系统平衡不均衡用户需求的能力,提升系统的服务水平和利润。In the vehicle-relocation problem of shared vehicles,the traditional threshold-relocation method based on inventory theory requires scheduling vehicles immediately when the number of vehicles in an area exceeds the threshold,which generates a large number of relocation tasks and reduces vehicle availability.Therefore,this paper proposes a vehicle adaptive centralized relocation optimization method in a simulation optimization framework for shared autonomous electric vehicle(SAEV)systems to improve the traditional threshold relocation method.Specifically,a discreteevent simulation model of the SAEV system is first constructed.The model divides the operation area with the station as the center,divides the user demand into intraregional user demand and interregional user demand,and achieves flexible service by automatically picking up and dropping off users by vehicles.Then,in the simulation model,a fixed time interval is set to solve a transportation problem based on regional thresholds,and a centralized vehicle relocation scheme is determined to replace the traditional threshold-triggered single-vehicle relocation scheme.Finally,the simulation optimization framework and the BO-SPSA algorithm are designed to efficiently solve the relocation threshold to maximize the daily profit of the SAEV system.The algorithm optimizes the parameters of the simultaneous perturbation stochastic approximation(SPSA)algorithm by Bayesian optimization(BO)to achieve a faster and efficient solution.The Chengdu case shows that(1)BO-SPSA can be solved more quickly and efficiently than other algorithms.(2)The small-scale and large-scale operation scenarios simultaneously verify that the adaptive centralized relocation strategy proposed in this paper can serve more users,reduce the number of scheduling trips,and obtain more operational profits than the traditional relocation strategy.(3)Regardless of the scale of the operation scenario,when the demand is small,the use of fast charging piles cannot effectively improve the profitability of th

关 键 词：智能交通 自适应调度 仿真优化 共享出行 自动驾驶汽车 贝叶斯优化 

分 类 号：U492.22[交通运输工程—交通运输规划与管理] TP18[交通运输工程—道路与铁道工程]