Application Of Genetic Algorithm In Optimized Control System For Vehicle Distribution

With the quick development of economic and society, the commoditycirculation is more and more. So more attention should be paid to thedevelopment oflogistics industry. Otherwise,logistics will restrict thequickdevelopment of economic to some extend. But at present, technology andmanagement level of logistics in Western area of China are all quitebackward. All kinds of accounts are at the stage of manual management. Inorder to change the backward situation, The Ministry of Communicationsestablished the project of the application research in distribution center inWestern area of China. The code of the project is 2004 398 81961.According to the study, the project is divided into six sub-problems.Designing and developing a logistics management information system is animportant one of the six parts. The use of the system can change thesituation that the logistics technologyand management methods of the Westare backward at present. Optimized Control System for Vehicle Distributionisapartofthewholesystem.Itisalsothekeyanddifficultyofthesystem.Firstly, through reading lots of relevant documents, all kinds ofalgorithms that solve vehicle routing problems are collected, arranged andstored. Based on this, I have thoroughly studied Genetic Algorithm (GA).When using GAto deal with vehicle routing problems, we often use ordinalstrings to code. Traditional GA using ordinal strings must use specialcrossover operators, such as PMX, OX, CX and so on. Such crossoveroperators not only are difficult to realize, but also need a diversity ofpopulation and easily convergence near by the optimal solution. Because ofthese, I firstly designed and implemented Partheno-Genetic Algorithm'sgene exchange operator, gene translocation operator and gene reverseoperator, then compared their effectiveness. Experimental results show that gene reverse operator has better performance. It not only has higherefficiency, but also has faster solving speed. In order to better effectivelyavoid the"Immature Convergence"problem, a Partheno-Genetic Algorithmwith a whistle is proposed. The algorithm sets up a whistle to preserve thebest individual that appeared on past dynasties. If the fitness of theindividual after gene reverse or gene mutate is higher than the fitness of theindividual preserved in the whistle, A copy of the individual will bepreserved in the whistle, which will speed up the convergence, therebyeffectively find the optimal solution. If the fitness of the individual aftergene reverse or gene mutate is lower than before, the individual after genereverse or gene mutate is also preserved in the group, which is of help toguarantee the diversity of the group, thus effectively avoid the"ImmatureConvergence"problem.ThismethodisdifferentfrommostotherTGAthosepreserve the better individuals, which easily form the"ImmatureConvergence"problem. Experimental results indicate that the improvedalgorithm in the article not only avoids effectively the common defects of"Immature Convergence"but also has better convergence property and findthe optimal or near-optimal solution effectively and quickly, which provideasolidtheoreticalfoundationforsolvingvehicleroutingproblems.Secondly, through investigating to the company that the projectsupports, I found that there exist some disparity between the theoreticalresearch and practical application. In theoretical research, each vehicle hasthesameloading.Allvehiclesstartofffromdistributioncenterandreturntodistribution center. But in practice, the vehicles in one scheduling generallyhave different loadings. One vehicle starts off from distribution center, butprobably return to another warehouse. In the aspect of time windows, thetheoretical research generally is hard time windows. When one vehiclearrives before the earliest service time of the customer, the vehicle can wait.But in practice, the time windows are wider at present. If the vehicle starts offatreasonabletime,itgenerallyneednotwaittocompletethedistributiontasks. And logistics companies hope the vehicles not wait. According topractice, the article established the mathematical model. Consideringflexibilityandpracticalityofthesystem,thepaperdesignedtwosolutionstosolvehardtimewindowsproblem.Onesolutionallowsvehiclestowait.Theother does not allow vehicles to wait. Both solutions allow users to flexiblyinput the number of the vehicles; each vehicle's loading, starting off place,returning place and so on. Users also can flexibly input each vehicle'sstarting off time in the first solution. After running, the system will outputevery vehicle's traveling distance, waiting time and routing information.The second solution does not allow inputting the vehicles'starting off time.But after running, the system will output each vehicle's permissible startingoff time windows, traveling distance and routing information. Users canbaseonthepractice,andthenflexiblyselectthebettersolution.Finally,thesystemwasrealizedontheJ2EEplatform.Andthesystemuses B/S architecture. Web application program uses Weblogic server. Thebackground of the system uses Sybase database. We carried on a realexample analysis to the practical problem and have finished the feasible testofthemethodandfunctiontest.At present, the system has basically realized. Testing work has beenfinished. And the system has entered trial running and debugging stage. Inthe period of trial running, the system has got better result in practice. Thecost for the logistics operation of enterprises has been cut down and theefficiencyhasbeenpromoted.