Loading Optimization Method Based On Two Level Clustering

With the development of globalization and outsourcing, logistics distribution in a large scale has become an important problem facing many enterprises, which processes the characteristics of dispersed distribution, abundant product categories, frequent undulation of customer demand, increasing in the situation of less-than-carload(LCL) freight transport. In order to cut down the cost of logistics,respond to market demand quickly and reduce the LCL situation, loading optimization and dispatching has attracted more and more attention.In this paper, a novel method called two level clustering using in resolving the problem of loading optimization has been proposed based on the mentioned background. By introducing an improved algorithm of Artificial ImmuneAlgorithm(AIA) with developing the ideas of this algorithm and systematic clustering the two level clustering method has been presented. The algorithm has the structure of, two layers. Firstly, implementing the upper level of the algorithm can achieve theeffect of grouping the customers into several clusters according to their positions. This stage has just taken the customer location into account without considering theneeds of theirs. Secondly, sort the orders according to the dispatching cycle, form a fleet remained to be assigned. Then, the dispatching plan calculated from the second level of the algorithm can be obtained. Therefore, based on the size of order, the diversity of product and the smallest split unit of product, several heuristic rules has been defined with consideration of using the minimal number of vehicles and allowing to split circumstances. The specific model has been designed as well. Finallyby combining the heuristic strategy and multi-level priorities strategy, the solution,, which takes the accuracy of delivery goods and the shortest route option into account, has been obtained. On the one hand, the algorithm of the upper level has been used to the calculation amount of the lower level. On the other hand, the model of lower levelwith heuristic rules and priority levels has been used to resolve the problem of each cluster. Thus, the presented algorithm can dramatically reduce the complexity and obtain the optimal solution of the problem. At last, simulation experiment has been carried out. The procedure and ideas have been illustrated to evaluate theeffectiveness of the algorithm.