Optimization of Network Coding Strategies for Multiple Information Flows

Abstract

Although the theory of optimal network coding for single-session multicast scenario has been well studied and applied, optimal network coding strategy for a more general case where there exists several sessions and receivers requesting for different information flows still needs more work. According to the recent study, some approaches, such as the pairwise-XOR strategy, are only feasible for multiple sessions that are restricted to very special topology structure. Otherwise, the processing time will be too long. In this project, firstly, a strategy named "selective random linear network coding", which is feasible for general connection problems, is introduced and implemented in Matlab. Secondly, to improve the performance, Immune Algorithm and Differential Evolution are applied. It is shown that the Immune Algorithm can help find the best solution faster but converges too fast to find the globally optimal solution. Differential Evolution, at the same time, can perform even better than the Immune Algorithm. In conclusion, the strategy introduced in this project is suitable for a general case and modification in evolution can help improve its performance.