Search Of Container Loading Methods Based On Estimation Of Distribution Algorithms

Container loading is an important process of goods transportation, since a reasonableloading plan can improve the utilization rate of container space and bring the companyprofit to the most by reducing operation cost. In theory, container loading problem is acomplicatedly combinatorial optimization problem with multi-constraints, which belongs tothe typical NP complete problem and difficult to obtain an optimal solution. As a result, itbecomes a really hard problem to improve the algorithms to solve the container loadingproblem in a better way.Estimation of distribution algorithm forms a new evolutionary pattern by combiningstatistical learning theory and evolutionary algorithm, which is a hotspot in the area ofevolutionary computation. This algorithm is a intelligent algorithm based on probabilisticmodel and has potential in solving all kinds of complicated problems and has already beenapplied in many areas and also made good achievements.This article introduces the related contents of the estimation of distribution algorithm,including the classification of the algorithm, theoretical research, current application anddevelopment trends; then presents the related concepts, mathematical model and constraintsabout container loading problem, introduces a variety of heuristic algorithms and intelligentalgorithm applied in the container loading problem and analyzes the existing researchresults.For the results of previous studies, this article proposed a new hybrid algorithm tocontainer loading problem by mixing estimation of distribution algorithm with heuristicalgorithm. Firstly, it locates and sorts good s to be loaded with Heuristic algorithms. Duringpractical operation, it sequence goods by descending sort the volume of goods and alsolocate goods by corner strategy. Secondly, it divides the free space by way ofthree-space-dividing and preserves the related data information under the first-in, last-outmechanism of stack. Lastly, it combines idle space with available space and make them intofully use. At the same time, it processes the data in free space by Multidimensional Arrays.The paper use estimation of distribution algorithms to process the population data bysampling and establishing the probability model to achieve population evolution and get theoptimal solution of container loading problem. The testing results illustrated the effectiveand existing deficiencies of the new algorithm.In the foundation of mixed algorithm, research in this article improves the UnivariateMarginal Distribution Algorithm，by way of keeping the elite population and adding in mutation operation, which brings out a new kind of mixed algorithm. Besides, moredetailed constrained conditions are considered in this new algorithm, such as containerloading capacity, container gravity limit and the setting direction of goods. The researchresults prove that compared with other current intelligent algorithms, the improved mixedalgorithm is more effective in solving complicated container loading problems, and canrealize its value in reality applications.