The solution to the capacitated vehicle routing problem (CVRP) is vital for optimizing logistics. However, the transformation of real-world logistics problems into the CVRP involves diverse constraints, interactions between various routes, and a balance between optimization performance and computation load. In this study, we propose a systematic model originating from the part logistic routing problem (PLRP), which is a two-dimensional loading capacitated pickup-and-delivery problem that considers time windows, multiple uses of vehicles, queuing, transit, and heterogeneous vehicles. The newly introduced queuing and transit complicate the problem, and to the best of our knowledge, it cannot be solved using existing methods or the standard commercial optimizer. Hence, this problem has caused the existing research to develop, generalize, and extend into the two-dimensional CVRP (2L-CVRP). To solve this problem, we provide a framework that decouples the combination of 2L-CVRP and global optimization engineering and derives an efficient and realistic solver that integrates diverse types of intelligent algorithms. These algorithms include: (1) a heuristic algorithm for initializing feasible solutions by imitating manual planning, (2) asynchronous simulated annealing (SA) and Tabu search (TS) algorithms to accelerate the optimization of global routes based on novel bundling mechanics, (3) dynamic programming for routing, (4) heuristic algorithms for packing, (5) simulators to review associated time-related constraints, and (6) truck-saving processes to promote the optimal solution and reduce the number of trucks. Moreover, the performances of the SA and TS solver algorithms are compared in terms of various size scales of data to obtain an empirical recommendation for selection. The proposed model successfully established an intelligent management system that can provide systematic solutions for logistics planning, resulting in higher performance and lower costs compared to that of manual planning.

有能力的车辆路径问题 (CVRP) 的解决方案对于优化物流至关重要。然而，将现实世界的物流问题转化为 CVRP 涉及多种约束、不同路线之间的相互作用以及优化性能和计算负载之间的平衡。在本研究中，我们提出了一个源自零件物流路由问题 (PLRP) 的系统模型，这是一个二维负载能力的接送问题，考虑了时间窗口、车辆的多次使用、排队、中转和异构车辆。新引入的排队和中转使问题复杂化，据我们所知，使用现有方法或标准商业优化器无法解决。因此，这个问题导致现有的研究发展、推广、并扩展到二维 CVRP (2L-CVRP)。为了解决这个问题，我们提供了一个框架，将 2L-CVRP 和全局优化工程的组合解耦，并衍生出一个集成了多种智能算法的高效、现实的求解器。这些算法包括：(1) 一种通过模仿人工规划来初始化可行解的启发式算法，(2) 异步模拟退火 (SA) 和禁忌搜索 (TS) 算法，以基于新颖的捆绑机制加速优化全局路线，(3 ) 路由动态编程，(4) 包装启发式算法，(5) 模拟器来审查相关的时间相关约束，以及 (6) 节省卡车的过程，以促进最佳解决方案并减少卡车数量。而且，SA 和 TS 求解器算法的性能在各种数据规模方面进行比较，以获得选择的经验建议。所提出的模型成功建立了一个智能管理系统，可以为物流规划提供系统的解决方案，与人工规划相比，具有更高的性能和更低的成本。