Optimization And Simulation Research Of Application Layer Multicast Tree Based On Memetic Algorithm

With the development of Internet, various multimedia businesses have emerged, such as telephone conference, the remote teaching, and TV conference. However, streaming media technology based on C/S style cannot meet the demand of people, especially when a large number of users are online meanwhile, the server performance will plummet. In order to solve the transmission problem of streaming media network, people put forward network multicast communication mode that a source node can send information to multiple destination nodes. Because of its own defect, such as the rapid expansion of routing table with users increasing, and the overweight of router load, network multicast technology has not been widely applied. In order to improve the efficiency of multicast, application layer multicast, which is a multi-user data distribution scheme based on application layer, has become the research hot spot.With the application layer multicast services are implemented, an important problem is the design of multicast tree. So the degree and delay constrained minimum spanning tree problem(DDCMST) is focused in this paper.(1) This paper proposed a new optimization algorithm to solve DDCMST problem using the basic idea of the Memetic algorithm, compares it with some other heuristic algorithms that were modified for the DDCMST problem. Memetic algorithmis a hybrid of an evolutionary algorithm with an individual learning procedure that has the capability of local improvement. The prominent advantage of Memetic algorithm is both global and local search in each iteration, therefore, the probability of finding the optimal solution is increased, and avoid the local optimum problem in great degree. Compared with Ant colony algorithm and genetic algorithm, simulation results show that Memetic algorithm is superior in efficiency.(2)We improve Memetic algorithm through combine rough search and fuzzy logic controller technique, new algorithm can improve the fitness of initial population and automatically adjust the parameters. Previous algorithm adjust the parameters depend on human experience and a large number of experiments for different topologies scale. The new algorithm can get rid of the problem caused by artificial selection. Simulation results show that the proposed method is feasible.