Linear Network Coding Construction for Multi-Source Multicast Network

The multi-source network coding problem, in which multiple multicast sessions with independent data share a network, is an open challenge. This paper proposes an approach to implement multi-source multicast sessions with linear network coding. We divide the original network into several sub-graphs, and each sub-graph contains a source node and its corresponding sink nodes. Links belonging to different sub-graphs do not overlap each other. In order to make the whole network throughput as large as possible, we transform partitioning sub-graphs into a combinatorial optimization problem and use genetic algorithm to solve it. Moreover, we give a pretreatment to narrow the search space. Then, it can be easy to construct coding scheme for each sub-graph. Furthermore, we prove that with the proposed approach, the throughput of the whole network is not less than the one with routing. Simulation results show that the proposed approach is feasible.