A New Approach to Network Optimization Using Chaos-Genetic Algorithm

Genetic Algorithms (GAs) have been widely used to solve network optimization problems with varying degrees of success. Part of the problem with GAs lies in the premature convergence when dealing with large-scale and complex problems; Caught in local optima, the algorithm might fail to reach the global optimum even after a large number of iterations. In order to overcome the problems with traditional GAs, a method is proposed to integrate Chaos Optimization Algorithms (COAs) with GA to fully exploit their respective searching advantages. The basic idea of COA is to transform the problem variables, by way of a map, from the solution space to a chaos space and to perform a search that benefits from the randomness, orderliness and ergodicity of chaos variable. In this chapter, we will first discuss network optimization in general, and then focus on how chaos theory can be incorporated into the GA in order to enhance its optimization capacities. We will also examine the efficiency of the proposed Chaos-Genetic algorithm in the context of two different types of network optimization problems, Grid scheduling and Network-on-Chip mapping problem.