An Improved Asynchronous Parallel Genetic Algorithm In Logistics Distribution Path

With the scale of the problem in the actual production and living growing increasingly complex solution space, more and more constraints, to find a common optimization method to solve complex optimization problems is very difficult. The emergence of Genetic Algorithm to solve the optimization problem provides the ideal solution, and its biggest feature is the implicit parallel computing features as well as the characteristics of the global search the entire solution space. However, Traditional Genetic Algorithm has the limitation and disadvantages. Therefore, in this paper, we bring forward an improved method, and apply it into the analysis of logistics distribution routing problem.In this paper, firstly, we introduce the basic related acknowledge and current research status of logistics distribution routing problem and Traditional Genetic Algorithm. Secondly, we analyze the advantage and disadvantage in Traditional Genetic Algorithm. Lastly, we focus on the main problem that taking too much times in calculation when Traditional Genetic Algorithm deal with the dramatic amount data, and proposing an Asynchronous Distributed Parallel Genetic Algorithm based on MPI and Parallel Genetic Algorithm. In the process of the algorithm running, the evolution of individual part in the population is asynchronous paralleled, which saved the time efficiently.In this paper, the proposed algorithm uses asynchronous parallel model for client-server mode. Each client to perform Genetic Algorithm for a subpopulation and the information of elite individuals in each generation sent to the server. The server is responsible for management the elite individual information for each client, and the evolution of subpopulations according to client migration. Of the proposed method in the migration phase migration only elite individuals, thus reducing traffic and time savings. Above on the basis of this article has designed a new fitness function, to sort the results of the fitness of each generation, for less than30percent of individuals will be adding a new fitness function of the penalty function its processing, thus ensuring the quality of the results of each generation.In this paper, we designed asynchronous parallel logistics path system, including the system’s function module design and simulation. The system function module is divided into four sections:System Settings module, user input module, asynchronous parallel genetic algorithm module and results output module. Then using simulated data of the system was tested. The experimental results show that the proposed algorithm is not only a very big advantage in time, and be able to get the ideal result of the operation.