Research On Optimal Construction Of Network Coding For Multi-source Multicast Connection

With the coming of information age, people demand a higher quality of network communication. Network coding technology is a major breakthrough in the field of network communication. Network coding based on multicast communication has become one of the important problems in communication research area.Aiming at multi-source multicast connection, this thesis proposes a network coding construction method to achieve the maximum vectors containing multicast capacities of sub-graphs, assuming that there is no ring. This method is bases on single-objective optimization to achieve the maximum throughput of the network. First it divides the network into several sub-graphs, which are single-source multicast networks. There are different groups which are made of single-source multicast networks in different division methods, and the multicast capacities constrain each other in the same group. In order to obtain the better division method to maximize the multicast capacity of each single-source multicast network, this paper uses particle swarm optimization algorithm to divide the network into sub-graphs, and solves the pareto solution set which contain the multicast capacity of each single-source multicast network dynamically. The pareto solution vectors in the pareto solution set correspond with the several optimalizing division methods. If users take into account the multicast capacity of certain single-source multicast network firstly, they can choose the right solution vector from the pareto solution set, then construct the linear network coding. In terms of improving the network throughput, the analysis shows that, the proposed approach is superior to traditional routing.Aiming at the directed cyclic network, this thesis proposes a construction method of network coding. First, it uses depth-first search to find out all the rings, and calculates the length of them, then removes one of the directed edges from each ring successively, and these edges do not participate the linear network coding construction. Thus, the directed cyclic network has been transformed into a directed acyclic network. There are different directed acyclic networks when different directed adges are removed. Finally, it achieves the throughput of each network, we can find out the directed acyclic network which has the maximum throughput and construct the linear network coding for it. Simulation and test results show that the proposed approach is feasible.