Research Of Delay-constrained Multicast Problem Based On Immune Clonal Computation And Multi-Agent Routing Algorithm

Multicast is a communicating process that one source node transmit the same information to lots of destination nodes (not all the nodes necessarily) in the network. Making use of this type of transmitting mode, data can be received by massive clients with the use of a little bandwidth so that information transmission will be more effective in the Internet. The key step of multicast communication is to select the multicast route, in other words, how to construct a multicast distribution tree to meet the QOS demands of users'. Therefore, it will make sense to guarantee multicast application's normal and efficient running by the multicast route that we found which not only meets the demand of QOS, but also with the smallest cost.The existing and general multicast algorithms are such as Flooding, DVMRP(Distance Vector Multicast Routing Protocol), and Layered Multicast Routing Algorithm, etc. However, these algorithms all have flaws. Flooding expends too much resource; DVMRP pruning has large efficiency, but it needs much space to save spanning tree. The use of Layered Multicast Routing Algorithm may cause the network state information inaccurate and delay because of layered treatment and regional classification.This paper put forward an algorithm to resolve the limited-delay multicast routing problem. The algorithm is designed for those nets which have high real-time demand. It adopted a heuristic idea based on the immune clonal algorithm combining with the multi-agent system。At first it searches for a alternative path set which meet the delay-constraint needs for each member in the multicast group, then use the cloning strategy and multi-agent system to evolve and select a better path to build the multicast tree.First, the paper expounds the research status of multicast routing technology and the theoretical foundation and algorithms of multicast tree. Then it studies the theoretical basis of multi-agent. Multi-agent system is a distributed autonomous system, it is a set of a great number of self-government-agent group. The paper researches the immune cloning calculating based on this, which is a betterment of genetic algorithm it is faster in convergence rate and stronger in search capabilities than the standard genetic algorithm. Then by combining this algorithm with the multi-agent systems, it gives a new way to resolve the delay-constraint multicast routing problem. The new algorithm first computes alternative paths set which meets the delay constraint for each multicast group members, and then uses the cloning strategy and multi-agent system to evolve and select a better alternative path to build the multicast tree. This algorithm uses prior knowledge of the issue to be solved to guide search and accelerate convergence, therefore avoids the problem of evolution blindness. In addition, the paper uses intelligent competition, cooperation, self-learning solution behavior to solve multicast routing problem, and achieves good results. Finally, the paper uses the simulation tool--NS-2 to practice the algorithm. According to the simulation result, the performance of the algorithm has been evaluated and the further improvement of it has also been pointed out.