A Study of the Effects of Clustering and Local Search on Radio Network Design: Evolutionary Computation Approaches

The goal of this paper is twofold. First, we want to make a study about how evolutionary computation techniques can efficiently solve the radio network design problem. For this goal we test several evolutionary computation techniques within the OPLINK experimental framework and compare them. Second, we propose a clustering approach and a 2-OPT in order to improve the results obtained by the evolutionary algorithms. Experiments carried out provide empirical evidence of how clustering-based techniques help in improving all algorithms tested. Extensive computational tests, including ones without clustering and 2-OPT, are performed with three evolutionary algorithms: genetic algorithms, memetic algorithms and chromosome appearance probability matrix algorithms.