The Best Planning And The Algorithm Research Of The Cargo Dispatch Of Physical Distribution Center

Nowadays, physical distribution has becoming the important element in promoting our country's economic development. At present, many physical distribution centers still use artificial measure and handle with problem with artificial managing experience; but this always make the cost higher and the efficiency lower. Here gradually comes a topic which is using informatization technology and computer on the management of physical distribution. "Informatization Physical Distribution" will gradually take up the main position.When the center is scheduling the goods, the two main problems are the scheduling of vehicle and the management of route. At present, the common way to solve these is genetic algorithm. When we use this algorithm, it is easy to get a best partial solution, but annealing algorithm is also a good way to get a globally best solution.Based on this point, the main work and the innovating points of this article are listed as below:This paper first introduced the basic acknowledge of physical distribution and the development of foreign and local computer management information center in the distribution center.Secondly, it established the dispatching model of goods and compare the various algorithm and the complexity analysis of it.Then I establish the genetic algorithm to the model and analyze the advantages and disadvantages of themselves from the two schemes. In the first scheme, we can use binary data as the code, Its advantages mainly reflects on finding out the best solutions in a short time and has an optimization effect on the fitness function and so on; the disadvantages mainly reflects on not considering about the secondary distribution about vehicles and being not adapted to the larger goods distribution. In the second scheme, we use natural number as the code, the main method is to generate a chromosome randomly, and adds the genes one by one, then judges whether these is sub-path when it takes part in, according to some certain stopping criterion. The main disadvantages are that it doesn't take an effective operation to the similar chromosome and may easily go into a local optimum solution.Fourthly, I established the simulated annealing algorithm, the main approach based on the thought of Metropolis Algorithm. Starting from an initial solution, ager several transformation of solution, it confirms a relative best solution based on a controlling parameter T. then we reduces the value of T and execute the Metropolis Algorithm circularly, when Tâ†'0), the overall optimal solution of combinatorial problem can be eventually found.This paper arouses a new algorithm --- genetic simulated annealing algorithm. It adds memory device, which imports two variable ii and ff, which is used to store the current best solution, ff to save its returning value. At first, we initialize ii as an initializing solution iiO, and ffo is taken as the corresponding function value. That is before in searching the new solution, we make ii=iiO, ff=ffO (ffO is decided by iiO), when it is in the cycle, every time we get a new solution (supposing it as ix) it will compare its corresponding target function value fx with the current function value ff. if fx is better than ff, we use ix and fx to instead ii an ff. at the end of the algorithm, it will compare the eventually best solution with the best one stored in the memory device, and at last takes the real best solution.Analysis of the article in the two experiments shows that the improved genetic algorithm stated in the article can significantly advance the speed of convergence of the algorithm firstly. Secondly, it can guarantee the solution is global optimum; at last, it takes the natural number as the coding scheme, so we can use small space to store larger customer information.It can also be seen from these two experiments that:Comparing the stated algorithm in the article with the traditional genetic algorithm, the advantages are that it can save computation time when it shortens the evolution algebra in order to get the same result in one hand; in the other hand, even if we take the same evolution algebra, although the improvement is not obvious when the evolution algebra is limited, with the increase of the evolution algebra, the result got by the algorithm stated in the article is obviously better than the one got by the traditional one.At the end of the article, I give a summary to the whole work I did in the article and an outlook to the next research work.