基于TD3的网联电动公交车生态驾驶优化方法  

A TD3-Based Eco-Driving Optimization Method for Connected Electric Buses

作　　者：郗毅 钱恒龙 潘应久 XI Yi;QIAN Henglong;PAN Yingjiu(School of Automobile,Chang’an University,Xi’an 710018,China)

机构地区：[1]长安大学汽车学院,西安710018

出　　处：《汽车工程学报》2025年第1期38-48,共11页Chinese Journal of Automotive Engineering

基　　金：国家自然科学基金项目(52402417);陕西省自然科学基础研究计划项目(2023-JC-QN-0385)。

摘　　要：针对纯电动公交车在信号交叉口高能耗的问题,提出了基于双延迟深度确定性策略梯度算法(TD3)的纯电动公交车信号交叉口生态驾驶优化方法。基于SUMO搭建仿真训练平台,综合考虑能耗、通行效率、舒适性和安全性,构建多目标优化的强化学习奖励函数;基于TD3深度学习框架,结合电动公交车在信号交叉口处的运行特征构建生态驾驶优化模型并进行参数训练;以信号交叉口经典通行策略GLOSA为基准方法,对提出的生态驾驶优化模型进行性能验证。结果表明,相较于GLOSA策略,提出的生态驾驶策略在信号交叉口4种典型场景下能耗分别降低了9.82%、26.13%、19.00%、14.51%,同时保证了车辆的安全、舒适和通行效率。To address the issue of high energy consumption in battery electric buses at signalized intersections,this paper proposes an eco-driving optimization model based on the Twin Delayed Deep Deterministic(TD3)policy gradient algorithm.First,a simulation training platform is developed using SUMO,which balances energy consumption,travel efficiency,comfort,and safety in a multi-objective optimized reinforcement learning reward function.Next,an eco-driving optimization model is created within the TD3 framework,tailored to the operational characteristics of electric buses at signalized intersections,and its parameters are trained.Finally,the performance of the proposed model is validated against the classic intersection passage strategy,Green Light Optimal Speed Advisory(GLOSA).The results show that the proposed eco-driving strategy reduces energy consumption by 9.82%,26.13%,19.00%and 14.51%in four typical intersection scenarios,while also maintaining vehicle safety,comfort,and travel efficiency.

关 键 词：生态驾驶策略 网联电动公交车 多目标优化 深度强化学习 信号交叉口 

分 类 号：U491[交通运输工程—交通运输规划与管理]