【作者】 李义华；

【导师】 李夏苗；

【作者基本信息】 中南大学 ， 交通运输工程， 2012， 博士

【摘要】 物流配送车辆调度是物流运输活动中的一个重要问题,也是国内外理论界研究的热点。目前,大多数研究集中在不同环境下车辆调度模型构建和车辆调度问题优化算法两个方面,相应成果丰富。动态、随机车辆调度问题也得到理论上的关注,越来越多的学者开始研究物流配送车辆调度智能决策问题。变约束、多扰动和强时变环境下的物流配送车辆调度决策方法及其智能决策系统方面存在着大量值得研究的问题。多智能体(Multi-Agent System, MAS)作为分布式人工智能研究的前沿领域和支持智能决策的重要方法之一,因其具备的特点而被公认为是研究各类复杂系统的重要理论模型。物流配送系统作为多目标复杂大系统,引入多智能体技术对实时物流配送车辆调度决策进行研究是一种新兴的解决方法,可以满足物流配送系统中车辆地理上分布、信息动态变化、顾客订单随时变化等众多要求,因此具备很好的理论意义和实用价值。本文对基于多智能体的物流配送车辆调度决策方法进行研究,主要内容和研究成果如下：(1)系统地对物流配送车辆调度有关理论和多智能体及其规划理论进行了归纳和总结。(2)将多智能体技术应用于配送企业建模和物流配送车辆调度决策,提出了宏观企业建模—中观业务流程建模—微观Agent模型设计的集成综合建模方法,并且利用该方法对物流配送车辆调度决策进行研究。(3)对涉及物流配送车辆调度决策的任务匹配方法以及多智能体自动协商两个关键技术进行研究。引进项目管理中的工作分解结构(Work Breakdown Structure, WBS)对物流配送车辆调度任务进行分解并提出任务分解策略,提出了一种信息不完全条件下的合作伙伴选择匹配方法；基于一种开放式异构多智能体协商框架,讨论了基于合同网的交互出价协商策略,对智能体自动协商进行研究,并构建算例对其进行分析和模拟。(4)采用双层规划的建模方法构造了基于业务外包的集装箱港口车辆调度问题数学模型,设计一种有效的禁忌邻域搜索算法对其进行求解；通过对实验结果进行统计分析,得出了一些有指导性意义的车辆调度业务外包规则。(5)采用面向智能体的系统开发方法,构建了物流配送车辆调度原型系统。阐述了多智能体原型系统的开发过程,对多智能体自动协商过程进行展示。结果表明,利用多智能体技术进行物流配送车辆调度决策是可行和有效的。更多还原

【Abstract】 Vehicle scheduling in logistics delivery is not only an important issue in the logistics and transport activities, but also a research focus of the theoretical circles at home and abroad. Recently, most studies focus on two aspects, one is vehicle scheduling model construction in different environment, the other is optimization algorithm of vehicle scheduling problem, and they have achieved rich results. Dynamic-random vehicle scheduling problem also has been a theoretical concern, more and more scholars start to study intelligent decision-making in vehicle scheduling in logistics delivery. There are a lot of content worthy to study under the condition of varying constraints, multi-disturbance and strong time-variable.As a research frontier in distributed artificial intelligence and also an important method supporting intelligent decision-making, multi-agent system has been recognized as an important theoretical model of various complex systems. As a complex multi-objective system, logistics delivery system based on multi-agent will be a new tool to deal with vehicle scheduling decision-making in real-time logistics delivery, it can satisfy the requirement of vehicle geographical distribution, information dynamic changes and customer orders changes in logistics delivery systems, so it has have an important theory significance and practical value.In this dissertation, the decision-making method of vehicle scheduling in logistics delivery based on multi-agent has been researched, the main content and achievements are as follows.(1) Theory of vehicle scheduling in logistics delivery and multi-agent planning theory are systematically summarized.(2) Multi-agent technology has been introduced into enterprise modeling and decision-making of vehicle scheduling in logistics delivery, an integrated modeling method of Macro Enterprise Modeling-Business Process Modeling-Microscopic Agent Designing has been developed, by this method, the decision-making of vehicle scheduling in logistics delivery is described and studied.(3) Two key techniques of task matching method and multi-agent auto-negotiation during the decision-making of vehicle scheduling in logistics delivery have been studied. The vehicle scheduling task has been decomposed by the introduction of the Work Breakdown Structure (WBS) and a task decomposition strategy has been proposed. Then, a new partner selection matching method under the condition of incomplete information is developed; an open heterogeneous multi-agent negotiation framework is constructed; the contract-net-based negotiation strategy of interactive policy is discussed and the auto-negotiation of agent is studied in detail, a case is constructed to analyze and simulate consequantely.(4) A mathematical model of outsourcing-based vehicle scheduling problem in container port is constructed by the method of Two-tier planning modeling, and an effective taboo search algorithm is designed to solve this mathematical model. By statistical analysis of the test result data, obtained a number of vehicle scheduling outsourcing rules which have guiding significance are developed.(5) A prototype system of vehicle scheduling in logistics delivery is constructed by the method of agent of programming. The development process of agent prototype system and the auto-negotiation process about agent are described. It is demonstrated that the study of decision-making of vehicle scheduling in logistics delivery by agent is not only feasible, but also effective.更多还原