基于双向协同Transformer的低碳多目标车辆路径优化

姚利军; 王可君; 邹庆奥; 邹暾

doi:10.19713/j.cnki.43-1423/u.T20251441

您当前的位置：

首页 >

文章列表页 >

基于双向协同Transformer的低碳多目标车辆路径优化

智慧交通与物流 | 更新时间：2026-07-24

- 基于双向协同Transformer的低碳多目标车辆路径优化
- Low-carbon multi-objective vehicle routing optimization based on bidirectional collaborative transformer
- 铁道科学与工程学报 2026年23卷第7期页码：3178-3191
- 作者机构：
  
  1.湖南省长株潭烟草物流有限责任公司信息部，湖南长沙 410119
  2.湖南工商大学智能工程与智能制造学院，湖南长沙 410083
  3.中国烟草总公司湖南省公司物流管理处，湖南长沙 410004
- 作者简介：
  
  邹暾(1972—)，男，湖南长沙人，高级工程师，从事人工智能、大数据、物流路径优化方面的研究；E-mail：45130393@qq.com
- 基金信息：
  
  湖南省烟草公司长沙市公司科技项目(CS2023KJ09)
- DOI：10.19713/j.cnki.43-1423/u.T20251441
  中图分类号： F252;U121
- 收稿：2025-09-15，
  
  网络首发：2026-07-24，
  
  纸质出版：2026-07-28
- 稿件说明：
移动端阅览
姚利军,王可君,邹庆奥等.基于双向协同Transformer的低碳多目标车辆路径优化[J].铁道科学与工程学报,2026,23(07):3178-3191.

YAO Lijun,WANG Kejun,ZOU Qing’ao,et al.Low-carbon multi-objective vehicle routing optimization based on bidirectional collaborative transformer[J].Journal of Railway Science and Engineering,2026,23(07):3178-3191.
姚利军,王可君,邹庆奥等.基于双向协同Transformer的低碳多目标车辆路径优化[J].铁道科学与工程学报,2026,23(07):3178-3191. DOI： 10.19713/j.cnki.43-1423/u.T20251441.

YAO Lijun,WANG Kejun,ZOU Qing’ao,et al.Low-carbon multi-objective vehicle routing optimization based on bidirectional collaborative transformer[J].Journal of Railway Science and Engineering,2026,23(07):3178-3191. DOI： 10.19713/j.cnki.43-1423/u.T20251441.

摘要

现有低碳车辆路径研究在同时取送货的多目标优化、客户满意度协调、语义与位置信息联合建模以及多车协同决策等方面，难以满足复杂城市配送场景下的高效与柔性调度需求。为此，本文构建了一种综合考虑动态交通状况与柔性服务时间窗口的多目标优化模型——多行程低碳车辆路径优化模型(TCSMOLCVRP)。该模型以最小化燃油消耗、碳排放、车辆运营、驾驶员雇佣以及延迟惩罚等综合成本为目标，将总运输成本与配送时间的联合优化作为目标函数，在满足时间窗、车辆容量和多行程等实际约束条件下，实现环境、经济与服务质量的多维协同优化。为高效求解该模型，设计了一种融合双向协同Transformer架构与深度强化学习的DACT-DRL算法。该算法通过共享双向协作注意力编码器提取节点与位置特征，利用信息交互机制增强上下文理解能力，并由节点选择解码器生成配送路径。同时引入近端策略优化(PPO)算法对策略网络进行训练与参数更新，从而在多目标框架下实现高效的路径搜索与决策优化。在Solomon标准数据集上的仿真实验表明，DACT-DRL算法在求解速度和成本控制方面均显著优于蚁群算法、禁忌搜索等传统启发式方法。特别是在软时间窗设置下，TCSMOLCVRP更贴近实际配送场景，综合运营成本降低25.56%，且求解时间较短，体现出更强的调度灵活性与更高的资源利用率。本研究为城市物流提供了兼顾经济效益与环保目标的智能优化方案，不仅拓展了车辆路径问题的建模与求解思路，也为深度学习与运筹优化融合的智能物流系统发展提供了新的技术路径与实践参考。

Abstract

The existing research on low-carbon vehicle routing still faces challenges in multi-objective optimization with simultaneous pickup and delivery

coordination of customer satisfaction

joint modeling of semantic and spatial information

and multi-vehicle collaborative decision-making. These limitations can hinder efficient and flexible scheduling in complex urban distribution scenarios. To address these issues

this study proposed a comprehensive multi-objective optimization model

which was the Traveling Cost and Service Minimization for Optimizing Low-Carbon Vehicle Routing with multiple trips (TCSMOLCVRP). It integrated dynamic traffic conditions and flexible service time windows. The model aimed to minimize the total comprehensive cost

including fuel consumption

carbon emissions

vehicle operation

driver employment

and delay penalties

by jointly optimizing total transportation cost and delivery time. Subject to practical constraints such as time windows

vehicle capacity

and multiple trips

the model enables multi-dimensional synergistic optimization across environmental

economic

and service quality objectives. To efficiently solve this model

we designed a novel algorithm

Dual-Attention Collaborative Transformer with Deep Reinforcement Learning (DACT-DRL). This algorithm employed a shared bidirectional collaborative attention encoder to extract node and positional features

enhanced contextual understanding through an information interaction mechanism

and generated delivery routes via a node-selection decoder. Furthermore

the Proximal Policy Optimization (PPO) algorithm was adopted to train and update the policy network

enabling efficient path search and decision-making within a multi-objective framework. Simulation experiments on the Solomon benchmark dataset demonstrate that the DACT-DRL algorithm significantly outperforms traditional heuristic methods such as Ant Colony Optimization and Tabu Search in both solution speed and cost control. Particularly under soft time window settings

the TCSMOLCVRP model better reflects real-world delivery scenarios

reducing total operational costs by 25.56% while maintaining short computation times. These results highlight can enhance scheduling flexibility and improved resource utilization. This research can provide an intelligent optimization framework for urban logistics that balances economic efficiency and environmental sustainability. It not only extends the modeling and solution methodologies for vehicle routing problems but also offers a novel technical pathway and practical reference for the integration of deep learning and operations research in the development of intelligent logistics systems.

关键词

Keywords

references

KUMAR R , UNNIKRISHNAN A , WALLER S T . Heuristic scheme for heterogeneous vehicle routing problem on trees based on generalized assignment and Bin-packing upper bounds [J ] . Transportation Research Record: Journal of the Transportation Research Board , 2012 , 2283 ( 1 ): 1 - 11 .

BANIAMERIAN A , BASHIRI M , TAVAKKOLI-MOGHADDAM R . Modified variable neighborhood search and genetic algorithm for profitable heterogeneous vehicle routing problem with cross-docking [J ] . Applied Soft Computing , 2019 , 75 : 441 - 460 .

周果 , 季彬 , 方晓平 . 多对多越库配送绿色车辆路径问题及算法研究 [J ] . 铁道科学与工程学报 , 2022 , 19 ( 8 ): 2202 - 2210 .

ZHOU Guo , JI Bin , FANG Xiaoping . Research on green vehicle routing problem and algorithm of multi-to-many cross-docking [J ] . Journal of Railway Science and Engineering , 2022 , 19 ( 8 ): 2202 - 2210 .

张佳荦 , 石小法 . 多车型冷链物流配送路径规划研究 [J ] . 交通与运输 , 2021 , 37 ( 6 ): 78 - 82 .

ZHANG Jialuo , SHI Xiaofa . Multi-type vehicle cold chain logistics distribution path planning [J ] . Traffic & Transportation , 2021 , 37 ( 6 ): 78 - 82 .

郭放 , 黄志红 , 黄卫来 . 考虑前置仓选址与服务策略的同时取送货车辆路径问题研究 [J ] . 系统工程理论与实践 , 2021 , 41 ( 4 ): 962 - 978 .

GUO Fang , HUANG Zhihong , HUANG Weilai . Integrated sustainable planning of fast-pick area network and vehicle routing with simultaneous delivery and pick-up [J ] . Systems Engineering-Theory & Practice , 2021 , 41 ( 4 ): 962 - 978 .

邱叶 , 胡军红 , 郭缙烨 , 等 . 不确定需求下考虑货物时间价值的多式联运路径选择方法 [J ] . 铁道科学与工程学报 , 2024 , 21 ( 3 ): 994 - 1003 .

QIU Ye , HU Junhong , GUO Jinye , et al . Multimodal transport path selection method considering time value of goods under uncertain demand [J ] . Journal of Railway Science and Engineering , 2024 , 21 ( 3 ): 994 - 1003 .

TAŞ D , DELLAERT N , VAN WOENSEL T , et al . Vehicle routing problem with stochastic travel times including soft time windows and service costs [J ] . Computers & Operations Research , 2013 , 40 ( 1 ): 214 - 224 .

张文会 , 郑文诏 , 向宇豪 , 等 . 基于深度强化学习的低碳运输路径优化模型 [J ] . 工业工程与管理 , 2025 , 30 ( 2 ): 169 - 177 .

ZHANG Wenhui , ZHENG Wenzhao , XIANG Yuhao , et al . A model for green vehicle routing optimization problem based on deep reinforcement learning [J ] . Industrial Engineering and Management , 2025 , 30 ( 2 ): 169 - 177 .

周鲜成 , 余玮瑛 , 李松明 , 等 . 考虑时间依赖和同时取送货的双目标绿色车辆路径问题研究 [J ] . 系统科学与数学 , 2024 , 44 ( 9 ): 2798 - 2815 .

ZHOU Xiancheng , YU Weiying , LI Songming , et al . Research on time dependent dual objective green vehicle routing problem with simultaneous pick-up and delivery [J ] . Journal of Systems Science and Mathematical Sciences , 2024 , 44 ( 9 ): 2798 - 2815 .

邓梦杰 , 李义华 , 徐贝 , 等 . 考虑碳排放和客户满意度的医药冷链物流配送路径优化 [J ] . 铁道科学与工程学报 , 2025 , 22 ( 5 ): 2100 - 2113 .

DENG Mengjie , LI Yihua , XU Bei , et al . Optimization of pharmaceutical cold chain logistics distribution path considering carbon emission and customer satisfaction [J ] . Journal of Railway Science and Engineering , 2025 , 22 ( 5 ): 2100 - 2113 .

邱金红 , 孙靖 , 仲兆满 . 基于配送收益均衡的多目标绿色车辆路径优化算法 [J ] . 控制与决策 , 2023 , 38 ( 2 ): 365 - 371 .

QIU Jinhong , SUN Jing , ZHONG Zhaoman . A multi-objective green vehicle routing optimization algorithm based on delivery benefit balance [J ] . Control and Decision , 2023 , 38 ( 2 ): 365 - 371 .

李国旗 , 郝志丹 , 杨佳鑫 , 等 . 运力短缺下的生活物资临时分配点选址-路径优化研究 [J ] . 交通运输系统工程与信息 , 2025 , 25 ( 2 ): 304 - 313 .

LI Guoqi , HAO Zhidan , YANG Jiaxin , et al . Living materials temporary distribution points location-routing optimization under transportation capacity shortage [J ] . Journal of Transportation Systems Engineering and Information Technology , 2025 , 25 ( 2 ): 304 - 313 .

张景玲 , 孙钰粟 , 赵燕伟 , 等 . 策略梯度的超启发算法求解带容量约束车辆路径问题 [J ] . 控制理论与应用 , 2024 , 41 ( 6 ): 1111 - 1122 .

ZHANG Jingling , SUN Yusu , ZHAO Yanwei , et al . Hyper-heuristic for the capacitated vehicle routing problem with policy gradient [J ] . Control Theory & Applications , 2024 , 41 ( 6 ): 1111 - 1122 .

罗佳 , 李朝锋 . 基于残差图卷积网络与深度强化学习的需求可拆分车辆路径优化算法 [J ] . 控制理论与应用 , 2024 , 41 ( 6 ): 1123 - 1136 .

LUO Jia , LI Chaofeng . The split delivery vehicle routing optimization with the residual graph convolutional network and deep reinforcement learning [J ] . Control Theory & Applications , 2024 , 41 ( 6 ): 1123 - 1136 .

唐开强 , 傅汇乔 , 刘佳生 , 等 . 基于深度强化学习的带约束车辆路径分层优化研究 [J ] . 系统工程与电子技术 , 2025 , 47 ( 3 ): 827 - 841 .

TANG Kaiqiang , FU Huiqiao , LIU Jiasheng , et al . Hierarchical optimization research of constrained vehicle routing based on deep reinforcement learning [J ] . Systems Engineering and Electronics , 2025 , 47 ( 3 ): 827 - 841 .

罗梦文 , 王恺 . 考虑车辆路径的多工厂生产与配送联合调度 [J ] . 运筹与管理 , 2024 , 33 ( 11 ): 51 - 57 .

LUO Mengwen , WANG Kai . Coordinated distributed hybrid flow shop scheduling and vehicle routing [J ] . Operations Research and Management Science , 2024 , 33 ( 11 ): 51 - 57 .

YU Fahong , CHEN Meijia , XIA Xiaoyun , et al . Logistics distribution route optimization with time windows based on multi-agent deep reinforcement learning [J ] . International Journal of Information Technologies and Systems Approach , 2024 , 17 ( 1 ): 1 - 23 .

HICKMAN A J . MEET: Methodology for calculating transport emissions and energy consumption [M ] . Office for Official Publications of the European Communities , 1999 .

HICKMAN J , HASSEL D , JOUMARD R , et al . Methodology for calculating transport emissions and energy consumption [J ] . http://www.researchgate.net/publication/310594252 , 1999 .

PALMER A . The development of an integrated routing and carbon dioxide emissions model for goods vehicles [D ] . Cranfield : Cranfield University , 2007 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于深度强化学习的大型高铁站到发线运用方案快速编制方法

不确定环境下应急物资协同储备与分配鲁棒优化

高速铁路协同智能选线关键技术体系与工程应用研究

交替极永磁同步电机优化设计及多物理场分析

硅酸钠改性与沙漠砂协同作用再生混凝土单轴抗压性能