Studying Different Variants of PBIL to Solve a Real-World FAP Problem in GSM Networks

In this paper we study different versions of the PBIL (population-based incremental learning) algorithm to evaluate and try to improve the results obtained by the standard version of the algorithm when it is used to solve a realistic-sized frequency assignment problem (FAP). PBIL is based on genetic algorithms and competitive learning, being a population evolution model based on probabilistic models. On the other hand, it is important to point out that frequency planning is a very important task for current GSM operators. The FAP problem consists in trying to minimize the number of interferences (or conflicts in the communications) caused when a limited number of frequencies has to be assigned to a quite high number of transceivers (and there are much more transceivers than frequencies). In the work presented here we take as initial point the results obtained with the standard version of PBIL and we perform on the one hand a complete study with six variations of the algorithm (PBIL-negativeLR, PBIL-different, etc.) and on the other hand a hybridization between PBIL and a local search method. Our final goal is to discover which approach can compute the most accurate frequency plans for real-world instances.