1 School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
2 School of Electronic Science and Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
Show less
文章历史+
Received
Published
Revised
2025-05-27
2026-06-28
2025-12-16
Online Date
2026-07-29
PDF (15560K)
摘要
为提升轨道物流系统的智能化与调度效率,提出了一种融合数字孪生技术与强化学习算法的轨道物流路径规划系统。首先,设计了一套层级分明、区域划分合理的控制系统架构,通过“分区域分层级”的混合控制网络,实现对物流机器人与轨道设施的高效协同管理。其次,构建符合ISO 23247标准的“四层五空间”数字孪生系统架构,确保物理系统与虚拟孪生体的高实时性与高一致性。针对多机器人路径冲突与路径规划效率问题,提出改进的双重深度Q网络算法(improved double deep Q-network,ID3QN),引入三维奖励机制(安全性、通行效率与时效性)与动作剪枝策略,有效提升路径规划的收敛性与稳定性。实验结果表明,该系统在复杂动态环境中展现出良好的调度性能,机器人平均路径时间缩短9.1%,冲突发生率降低12.3%。相关仿真与实验测试工作验证了所提方法在实际轨道物流场景中的可行性,体现了其在提升轨道物流系统智能化水平方面的应用价值。
Abstract
To enhance the intelligence and scheduling efficiency of rail-guided logistics systems, a path planning system integrating digital twin technology with reinforcement learning algorithms was proposed. First, a hierarchically structured and regionally segmented control system architecture was designed. By adopting a hybrid control network based on “regional division and hierarchical management,” the system enables efficient collaborative control of logistics robots and track-side facilities. Second, a “four-layer, five-space” digital twin architecture conforming to the ISO 23247 standard was constructed to ensure high real-time responsiveness and consistency between the physical system and its virtual counterpart. To address path conflicts among multiple robots and improve planning efficiency, ID3QN (improved double deep Q-network) algorithm was proposed. This algorithm introduces a three-dimensional reward mechanism—comprising safety, traffic efficiency, and timeliness—along with an action pruning strategy, significantly enhancing the convergence and stability of path planning. Experimental results demonstrate that the proposed system exhibits excellent scheduling performance in complex and dynamic environments, reducing the average robot path time by 9.1% and lowering the conflict rate by 12.3%. Relevant simulation and experimental testing have validated the feasibility of the proposed method in practical rail-guided logistics scenarios, demonstrating its application value in enhancing the intelligence level of rail-guided logistics systems.
针对上述问题,现提出一种以数字孪生驱动的轨道物流融合分层控制系统,并引入改进型双重深度Q网络(improved double deep Q-network,ID3QN)算法进行路径规划优化。具体工作包括:①设计符合ISO 23247[18]标准的“四层五空间”数字孪生系统架构,构建由接入层、汇聚层、核心层与用户层构成的分布式虚实融合体系;②提出“分区域+分设备”的混合控制结构,实现固定轨道设备与移动机器人控制方式的动态协同,提升系统扩展性与通信冗余能力;③基于D3QN算法引入初始经验、动作剪枝与冲突消解机制,形成多机器人协同路径规划的ID3QN算法,提升规划效率与系统稳定性,为轨道物流系统在复杂环境下的多机器人协同调度提供一种可行的架构设计与算法方案,对智慧医院、智能制造等场景下室内物流系统的实际部署具有一定的工程参考价值。
轨道物流调度系统与其他智能制造领域的数字孪生系统一样,面临设备结构复杂、数据量庞大、实时性要求高等挑战。为应对上述问题,已有诸多学者提出“分布式数字孪生”[20]概念,其核心思想是在系统中构建多终端、多模型的协同交互体系,将计算任务分散至多个节点,并通过网络实现高效协同,从而提升数据处理效率与系统响应能力。基于上述理念,参考国际标准化组织(International Organization for Standardization, ISO)发布的制造系统数字孪生标准体系 ISO 23247,融合分布式数字孪生架构思想,面向轨道物流应用场景,设计了一种“四层五空间”数字孪生系统架构,如图3所示。
在轨道物流机器人接受任务调度后,如何选择一条时间最短,路程最近,少拥堵,无冲突的最优路径,是需要解决的核心问题。因此,在D3QN(dueling double deep Q-network, D3QN)算法上提出了ID3QN(improved dueling double deep Q-network, ID3QN)算法,设计了三维评价指标与动作剪枝,并将其作为奖励函数,从而得到最优的路径规划。与传统单机器人启发式算法不同,深度强化学习在解决大规模机器人配送场景拥有着出色的表现。将各个物流机器人看作一个个独立的智能决策数字孪生个体,在核心层数字孪生空间内进行探索学习,通过奖励函数与孪生空间环境进行交互,从而在真实场景中进行决策。
KangZhibang. Design and development of medical logistics rail transit system[D]. Chengdu: University of Electronic Science and Technology of China, 2024.
[3]
LengK, LiS. Distribution path optimization for intelligent logistics vehicles of urban rail transportation using VRP optimization model[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(2): 1661-1669.
CaiQiwen, MaJidong. Optimization of urban low-carbon logistics path based on BSO_ALNS algorithm[J]. Science Technology and Engineering, 2025, 25(9): 3921-3930.
[6]
JiangY H, GaoS, YinY H, et al. A control system of rail-guided vehicle assisted by transdifferentiation strategy of lower organisms[J]. Engineering Applications of Artificial Intelligence, 2023, 123: 106353.
[7]
FangB, FengM, QiuC, et al. Rail multi vehicle scheduling method for intermediate depot of steel plant based on improved genetic algorithm[J]. Scientific Programming, 2022, 2022(1): 4971638.
ZengAijing, LiuYongjiang, ChenYuepeng, et al. Optimization of logistics distribution scheduling based on digital twins[J]. Science Technology and Engineering, 2021, 21(21): 9005-9011.
[10]
LiuW, WuM, WanG, et al. Digital twin of space environment: development, challenges, applications, and future outlook[J]. Remote Sensing, 2024, 16(16): 3023.
JiaDongli, KangTianyuan, WangShuai, et al. State prediction method for digital twins of power distribution networks by integrating temperature factors with wavelet LSTM[J]. Science Technology and Engineering, 2025, 25(7): 2817-2824.
[13]
ZhangD, LiuZ, LiF, et al. The rapid construction method of the digital twin polymorphic model for discrete manufacturing workshop[J]. Robotics and Computer-Integrated Manufacturing, 2023, 84: 102600.
YanQiong, ZhangHaijun. Quality evaluation of digital twin workshop based on hesitant fuzzy linguistic[J]. Science Technology and Engineering, 2020, 20(16): 6472-6478.
[16]
KaewunruenS, LianQ. Digital twin aided sustainability-based lifecycle management for railway turnout systems[J]. Journal of Cleaner Production, 2019, 228: 1537-1551.
[17]
LiuS, ShiZ, LinJ, et al. A generalisable tool path planning strategy for free-form sheet metal stamping through deep reinforcement and supervised learning[J]. Journal of Intelligent Manufacturing, 2025, 36(4): 2601-2627.
[18]
HanH G, XuZ A, WangJ J. Reinforcement learning-assisted particle swarm algorithm for effluent scheduling problem with an influent estimation of WWTP[J]. Swarm and Evolutionary Computation, 2025, 94: 101871.
[19]
AbdulsahebJ A, KadhimD J. Classical and heuristic approaches for mobile robot path planning: a survey[J]. Robotics, 2023, 12(4): 93-114.
FanHouming, MuShuang, YueLijun. Collaborative optimization of automated guided vehicle scheduling and path planning consi-dering conflict and congestion[J]. Journal of Computer Applications, 2022, 42(7): 2281-2291.
[22]
NishidaK, NishiT. Dynamic optimization of conflict-free routing of automated guided vehicles for just-in-time delivery[J]. IEEE Transactions on Automation Science and Engineering, 2022, 20(3): 2099-2114.
ZhangKaiyue, WenHaijun, ChenYuepeng, et al. Vehicle transport path optimization method of hazardous chemicals based on digital twin technology[J]. Science Technology and Engineering, 2023, 23(34): 14676-14681.
[25]
PavlosE, StavrosL, StathisP, et al. Cognitive digital twins for resilience in production: a conceptual framework[J]. Information, 2022, 13(1): 33.
[26]
余鹏. 多元耦合下主动配电网集中-分布协同调控研究[D]. 杭州: 浙江大学, 2023.
[27]
YuPeng. Research on centralized-distributed coordinated control of active distribution networks under multivariate coupling[D]. Hangzhou: Zhejiang University, 2023.
[28]
LiuZ, ZhangY, YangC, et al. Generalized distributed four-domain digital twin system for intelligent manufacturing[J]. Journal of Central South University, 2022, 29(1): 209-225.