A Hybrid Genetic Algorithm for the Fixed Channel Assignment Problem

This paper describes a hybrid genetic algorithm for solving instances of the Fixed Channel Assignment Problem (FCAP), a problem that is frequently encountered by designers of mobile telecommunication networks. The hybrid GA manipulates solutions which model networks directly, allowing it to provide realistic assignments for highly constrained problems. Unfortunately, such solutions can be very expensive to evaluate. Algorithms such as simulated annealing often speed up the evaluation process by using delta evaluation. Whilst such an approach is not normally adopted by genetic algorithms, this paper demonstrates that delta evaluation can be incorporated into a GA, to deliver dramatic speed increases. We have found that delta evaluation can improve the speed of our GA by a factor of 90. This improved performance allows the GA to produce good results for large and complicated networks in a reasonable amount of time. The results obtained by the GA are compared to previous GA algorithms proposed for the FCAP and to a highly tuned simulated annealing algorithm.