论文标题:实时订货信息下物流配送车辆调度优化研究
Research on the Vehicle Scheduling Optimization of the Logistics Distribution with Real-Time Demand Information
论文作者 柳伍生
论文导师 胡列格,论文学位 硕士,论文专业 交通运输规划与管理
论文单位 长沙理工大学,点击次数 108,论文页数 93页File Size4231k
2004-04-20论文网 http://www.lw23.com/lunwen_423157357/ 电子商务;实时订货信息;物流配送;车辆调度优化;时间窗;混合遗传算法;禁忌搜索法
The electronic commerce;Real-time demand information;Physical distribution; Vehicle scheduling problem;Time windows; Hybrid genetic algorithm; Tabu search algorithm.
由于电子商务和网际网络的兴起,越来越多的企业开始应用电子商务和网络,同时顾客对商品到货时间要求越来越严格。可以说,对顾客进行产品的及时配送是企业和电子商务成功的关键,而适当的车辆调度方式是较少配送时间和配送成本的重要因素。 在过去,对实时订货信息下车辆调度优化的研究,只能将这些实时信息累积,待各车辆服务完所有预定的顾客后再重新进行配送,一旦车辆路线决定并了进行配送,在获得新的信息后将无法更改,而只能新增加车辆进行配送或者在途车辆原路返回补货后再按原配送路线配送。事实上,这种调度方式已经失去了实时订货信息所具备的优势,忽视了需求信息的改变对最佳配送路线的影响。它一方面可能导致无法有效满足顾客的要求,另一方面,也可能大大增加配送中心的配送成本。因此,传统的车辆调度问题算法已无法应付快速回应顾客需求以及配送中心对物流配送提出的要求。 本研究尝试利用动态的观点处理实时订货信息下具有随机性需求量与需求地点的车辆优化配送问题,期望在订货需求信息不断变动的情况下,适时改变车辆配送的路线与增加车辆进行服务,使其能够更有效率的服务所有需求。为此,本文探讨了电子商务出现后,实时订货信息对于物流配送的影响,并分析电子商务下的配送与传统配送不同特性。由此构建出符合此物流配送特性的问题模式,并结合实时的订货信息与车辆派遣,建立一套能够处理该模式的车辆调度方法。 在模型构建中,运用处理静态车辆调度问题的车辆运营变动成本模型,在此基础上,充分考虑顾客对时间要求和实际配送特性,加入混合的时间窗模型,并考虑本研究环境下,可能发生配送失败情形,配以配送失败的惩罚成本,构建符合本问题的模式。并依据本研究的研究范围和假设,设计出适用于本研究的演算流程:初始路线构建和路线改善。初始路线将遗传算法全局搜索能力强的特点和局部搜索算法局部搜索能力强特点相结合,构建了适于本文的混合遗传算法,路线改善采用禁忌搜索法,经案例测试,该方法在求解时间和效果上表现出良好的性能,尤其是在求解大规模的车辆调度问题,具有一定的实际应用价值。
With the rising of the electronic commerce and network, more and more firms have devoted themselves in the application of the Internet network on their businesses and the customers are more and more stringent to time for the goods delivery. The capability of delivery their products to the customers in the shortest time is the key factor of the success of the electronic commerce. The appropriate delivery route assignment plays major role in the reduction of the delivery operation time.Traditionally, the studies about Vehicle Scheduling Problem (VSP) on real-time demand infromation can distribute only after get together demand information and waiting for all vehicles having arrived. So once the route structure is determined, it will not be altered. If distribution center will ment with demand of customers, they often add vehicles to service or order vechicles returning to distribution center to supply goods.In fact, they couldn"t take advantage to the advantage of real-time information well. This type of approach ignores the impacts of the changes in demand requirements and traffic conditions on the optimal route structure, on the one hand,it may lead to distribution center not to ment with demand of customers,on the other hand,it may add costs of distribution significantly. Therefore, the past method cann"t satisfy with the demand of the customers and distribution centers for the logistics distribution.The VSP with stochastic demand locus and quantities on the real-time demand information was dealt with in this paper, expecting to adjust vehicles routing and add vehicles to serve customers so that they can serve more efficiently. Therefore, the affect of real-time demand information on distribution was discussed in the paper and the difference between real-time distribution and traditional distribution was analyzed. Based on it, the paper built a corresponding mathematic model and designed a suit of method to solve the problem.In building the model, the paper applies to vehicle changing cost of static state"s VSP, and adds to hybrid time windows model according to demand and distribution characteristic in our country. Taking into account possible fail to distribute under the conditions on this paper, the punish cost was added when failing to distribute in the paper, thus, the three parts made up of the modle. By the confine, this paper designed a method to solve the problem: structuring initialization route and ameliorating route. On the basis of analyzing the weakness of genetic algorithm in local search, this paper builds a hybrid genetic algorithm which is the combination of genetic algorithm and local search algorithm for solving physical distribution routing problem. When ameliorating route, it adopted Tabu Search Algorithm, examples performs well in both
|
