物流业是关系国家经济发展的血脉，我国十三五规划纲要进一步从国民经济发展各个层面规划了现代物流业发展的重点及要点。而车辆路径问题作为物流配送系统优化的主要环节和因素，为了进一步降低配送成本和提高客户满意度，结合城市环境下物流车辆调度运行的基本状况，本文以基于群体智能优化算法的多行程车辆路径问题作为主要研究对象，分别研究了：(1)采用改进烟花算法来优化物流配送过程中多行程车辆路径的调度；(2)利用混合Beam-PSO优化算法来求解考虑货物到达时间带时间窗的多行程车辆路径问题；(3)利用混合教与学优化算法来求解运送不相容货物的带时间窗的多行程车辆路径问题： (4)利用新型混合蛙跳算法来同时进行配送中心选址定位和配送车辆行驶路线规划。最后，通过一个车辆调度与优化平台对本文提出的群体智能优化算法进行了实例验证。主要研究工作如下： 首先说明了所要研究课题的来源及研究目的，介绍了车辆路径优化问题的定义、模型及基本构成要素，并引入了动态需求VRP的求解策略和相关算法。在介绍国内外城市物流配送的基本问题及其特征的同时，即配送的动态性、多层次、多行程，也把基于群体智能算法和动态需求的VRP问题融入到了本文的研究内容中。最后指出本文的主要研究工作及组织结构。 针对基于改进烟花算法的多行程车辆路径问题，在标准烟花算法的基础上，构建了一种改进烟花算法。该算法结合多行程车辆路径问题的性质，采用了基于轮盘赌的编码与解码方法，克服了烟花算法无法适用于离散问题的缺点；同时，利用蜂群搜索技术进一步强化标准烟花算法的深度开发能力和搜索性能，并在该算法中融入了基于反学习的初始化方法用于提升初始解的质量。并分析对比了该算法的复杂度。该算法通过测试算例的参数设置和参数校正；又与文献中提出的遗传算法、人工蜂群算法和粒子群算法进行对比获得较好的结果，在MATLAB仿真实验中验证了该算法的可行性和有效性，证明了该算法在求解多行程车辆路径问题时能够获得最短的配送时间，总时间偏差百分比能够达到5％以上。 针对考虑货物到达时间的带时间窗的多行程车辆路径问题，提出了一种混合Beam-PSO优化算法来进行求解。首先借助随机键编码机制，构建了相应的编码与解码方法，克服了标准PSO算法无法适用于离散问题的缺点；同时引入了基于Beam Search的局部搜索流程以强化算法的全局搜索能力，并分析对比了该算法的复杂度。这里假定客户位置节点、客户需求量、到货时间和客户服务时间都是随机参数，最后通过MATLAB仿真实验，验证了该Beam-PSO优化算法在求解该问题时，相比人工蜂群算法、遗传算法和教与学优化算法可以获得最少的行程数量、最短的配送时间和最低的运输成本。 为了解决运送不相容货物的带时间窗的多行程车辆路径问题，需要制定一个明确的路径规划来服务一组客户，满足客户运送不相容的大宗货物的需求。设计了一种混合教与学优化算法，并引入了客户位置节点、客户需求量、客户服务时间、客户类型、货车容量以及不同货物的单位运输成本等动态因素。该算法通过借助随机键编码机制，克服了标准教与学优化算法无法适用于离散问题的缺点；构造了基于禁忌搜索算法的局部优化方差，进一步强化标准TLBO算法的寻优能力，并分析对比了该算法的复杂度。为了测试提出的混合教与学优化算法，分别同人工蜂群算法、标准粒子群算法和入侵杂草优化算法相比较， MATLAB仿真结果验证了这种混合算法在求解该问题方面可以获得最少的行程数、最短的配送时间和最低的运输成本。 为了同时解决多行程车辆路径问题和配送中心的选址定位问题，本文首先开发了一个以最小化总成本为目标的数学模型，其中包括运输成本和启动车辆成本。然后采用了新型混合蛙跳算法解决这个问题，设计了该问题的编码与解码方法，并进一步借鉴入侵杂草优化算法的邻域搜索模型，对各个青蛙子群体中的最优解进行局部搜索，强化了算法的进化能力。并分析对比了该算法的复杂度。这种方法不同于常规的把配送中心选址和路径优化分为两个独立阶段求解LRP的方法，而是考虑了配送中心设施选址和车辆路径两个问题之间的相互影响和制约关系，把LRP看做一个整体进行结构分析和求解。然后采用MATLAB进行实例仿真，分别与标准粒子群算法、人工蜂群算法和遗传算法进行对比实验。最后针对大规模客户点问题进行了理论探讨，并通过仿真进行了验证，结果表明该新型混合蛙跳算法可以最低的目标函数成本和最少的配送中心数量更好地获得配送中心选址和有效的路线安排，相比其他三种算法实用性更强，寻优效率更高。 最后介绍了一个实例---基于动态需求的车辆调度与路径优化平台，该实例有效集成了本文所介绍的改进烟花算法、混合Beam-PSO优化算法、混合教与学优化算法和新型混合蛙跳算法，并将它们作为优化种子通过GIS接口植入系统的SDK算法包中。针对调度过程中出现的不确定因素，如客户位置节点、客户需求量、客户服务时间等，该平台采用EXCEL表导入实时的更新数据并采用重新优化的策略，以三辆配送车为例验证了当前车辆行驶路线的优化过程，并把平台优化后的路径结果由调度人员决策选择后，通过短信发送给司机，实现动态调度的过程。 在全文的结论部分，对上述基于群体智能优化算法的多行程车辆路径问题有关的研究工作做了全面的总结，提出了本文研究的创新点，并指明了以后研究的方向。 以上理论成果对城市物流环境下多行程车辆路径问题的研究具有一定的指导意义，有助于提高城市配送系统的效率，具有良好的社会和经济效益。 关键词：群体智能优化算法；车辆路径问题；多行程；动态需求

 Logistics industry is the bloodline about the development of national economy, the 13〓 five year planning outline of our country further programmes the development key points of modern logistics industry from all levels of the national economic development As the key part of logistics system optimization, combined with the basic situation of logistics vehicles dispatching operation in urban area, this paper regarded the multi-trip vehicle routing problem based on swarm intelligence optimization algorithm as the main research object respectively studied the followings: (1) using an modified firework algorithm to optimize the scheduling of multi-trip vehicle routing in logistics distribution. (2) using a hybrid Beam-PSO algorithm to solve the multi-trip vehicle routing problem with time window and the release time based on dynamic demand; (3) using a hybrid teaching and learning optimization algorithm to slove mutli-trip vehicle routing. problem with time windows to delivery incompatible goods; (4) using a novel shuffled frog leaping algorithm to optimize the location of the distribution center and the effective route planning of the vehicle. Finally via a detail vehicle dispatching and optimization platform of company, the proposed the swarm intelligence algorithm is verified. The main research content is as follows: At first the source of the problem and research purposes are introduced, the definition, model and basic elements of vehicle routing optimization are discussed, and the solving strategy and related algorithms of VRP based on dynamic demand is introduced. At the same time the basic problems and characteristics of urban logistics domestic and overseas are introduced, that is, time-dependency, multi-level, multi-trip and dynamic information. Also the VRP based on swarm intelligence algorithm and dynamic demand is integrated into the research. Finally the technical route and organization of the main research work in this paper are pointed out. Aiming to the multi-trip vehicle routing problem, this section proposes a modified firework algorithm based on the standard firework algorithm. This algorithm combines the nature of multi-trip VRP, adopts the encoding and decoding method of roulette to overcome the shortcomings of the standard firework algorithm which are not applicable to discrete problem. At the same time, using bee colony search technology to further strengthen the depth development capability and search performance of the standard firework algorithm, and embodied in the firework algorithm with initialization method based on unlearning to improve the quality of the initial solution, and the complexity of the algorithm is analyzed and compared. Through the parameter setting and parameter correction, in the MATLAB simulation the algorithm's reliability and feasibility are verified. The algorithm is tested comparing with genetic algorithm, artificial bee colony algorithm, and particle swarm optimization algorithm in the literatures, encouraging results are obtained. Finally it is proved that this algorithm can obtain the shortest delivery time when solving the problem of multi-trip vehicle routing, and the total percentage of time deviation can reach above 5%. Aiming to solve the problem of the multi-trip VRP with time windows and release date based on dynamic demand, this section proposes a kind of hybrid Beam-PSO optimization algorithm to solve. At first this section constructs the corresponding encoding and decoding method via the random key coding mechanism, overcomes the shortcomings of standard PSO algorithm which is not applicable to the discrete problem. At the same time the local search procedure based on Beam Search is introduced to enhance the global search capability of the algorithm, and the complexity of the algorithm is analyzed and compared. Here the customer location, quantity demanded, release date and customer service time are all supposed random factors, finally through the MATLAB simulation, comparing with the artificial bee colony algorithm, genetic algorithm, as well as teaching-learning algorithm in related literatures, the experiments prove this Beam-PSO algorithm can gain the least amount of trips, the shortest delivery time and the minimum transportation cost. Aiming to solve the multi-trip vehicle routing problem with time windows and incompatible commodities, it is necessary to make a clear routing planning to serve a set of customers that meets the requirement of customer carrying large amount of commodities. This section design a hybrid teaching and learning optimization algorithm, and introduced the dynamic factors such as customer location, quantity of requirement, service time, customer type, and load capacity. This algorithm overcomes the shortcomings of standard teaching-learning algorithm which is not applicable to the discrete problem, constructs the local optimization variance based on Tabu search algorithm; further enhances the optimization capability of standard teaching-learning optimization algorithm, and the complexity of the algorithm is analyzed and compared. In order to test the proposed hybrid teaching-learning algorithm, comparing with artificial bee colony algorithm, particle swarm optimization algorithm, and invasive weed optimization algorithm, the MATLAB simulation results verify that the hybrid algorithm can obtain the minimum number of trips, the shortest delivery time and the lowest transportation cost in solving the problem. In order to solve both the multi-trip vehicle routing problem and the distribution center location problem at the same time, a mathematic model is developed at first. The objective of this model is to minimize the total costs including the transportation costs and the activated vehicle costs. A novel shuffled frog leaping algorithm is then developed for solving the problem. The coding and decoding method is designed, the neighborhood search model of invasion weeds optimization algorithm is further referred, and locally searching for the optimal solution of the frog subgroups, the evolution ability of algorithm is strengthened. This method is completely different with the common way of dividing distribution center location and path optimization into two separate stages to solve LRP method, but considering the mutual influence and restriction relationship between the facility location of distribution center and vehicle routing problem, the LRP is regarded as an entirety to do structural analysis and solving. Then MATLAB simulation is respectively done for the contrast experiment with other three kinds of optimization algorithm. Finally, this paper discusses the problem of large scale customer point in theory and verifies it through simulation. The results show that the novel shuffled frog leaping algorithm can obtain better distribution center location and effective route arrangement with the lowest cost of objective function and the least number of distribution centers, comparing with other three algorithms its practicability is stronger, and the optimization efficiency is higher. Finally a vehicle scheduling and routing optimization platform based on the dynamic demand is presented, the instance effectively integrates the modified firework algorithm, the hybrid Beam-PSO algorithm, a hybrid teaching and learning optimization algorithm and a novel shuffled frog leaping algorithm presented in the paper, and takes them as optimization seeds imbedded into the SDK algorithm package by GIS interface. For the uncertain factors in the process of scheduling, the platform adopts the strategy of real-time update of data and re-optimization, taking three delivery vehicles as examples, the optimization process of the current vehicle driving route is verified. Finally the optimized path results by scheduling personnel selection and decisions, via text message are sent to the drivers, to realize the process of dynamic scheduling. In the conclusion of the whole study, this section makes a comprehensive summary to the above research content, proposes the main innovative points, and points out the next research direction. Above theoretical results have certain guiding significance to the multi-trip vehicle routing problem in urban logistics environment, are helpful to improve the efficiency of urban distribution system, with good social and economic benefits. Key words: Swarm intelligence optimization algorithm; Vehicle routing problem;Multi-trip; Dynamic demand