A genetic algorithm for the channel assignment problem

This paper considers a genetic algorithm approach to the channel or frequency assignment problem in cellular telephone networks. This NP-hard problem is commonly solved using heuristics such as simulated annealing, neural networks and genetic algorithms. Previous genetic algorithm approaches, however, have had problems generating solutions which are both optimal and feasible simultaneously. In this paper, we use an alternative representation of the solution which, together with appropriate definitions of the crossover and mutation operators, enables feasibility of the solutions to be guaranteed. Furthermore, our definition of these operators helps to provide insights into the unique roles that the play in the optimisation process. The task of selecting parameters is then made easier through the isolation of these operators and by understanding the contribution each makes to the final solution