Genetic algorithm solution of network coding optimization

Abstract

After the best optimizing approach of network coding is being studied, some methods are proposed based on the characteristics of the network coding overhead optimization problem. First, two modifications are added to the preprocessing phase: 1) How to generate a fitness value to a network coding scheme under a certain network coding optimization request is presented. This makes different network coding optimization problems be solved with the same genetic algorithm. 2) An additional exam processing of the multi-in outgoing links is imported to reduce the solution space. Second, experimental results show that the random generated solution of network coding optimization problem can hardly achieve the multicast rate, three new steps are suggested be taken with the common genetic algorithm: 1) use more delicate member generating function to generate initial members; 2) add new members at the beginning of each round of the genetic algorithm to avoid localized optimization; 3) assign a fitness value based on each receiver's data rate rather than -1 to those network coding solutions which cannot achieve the max multicast rate. Experimental results show dramatic improvements in terms of both efficiency and result. © by Institute of Software, the Chinese Academy of Sciences. All rights reserved.