Wireless Network Gateway Placement by Evolutionary Graph Clustering

Abstract

Gateway placement is an important problem in the design of a backbone wireless network (BWN) as it directly affects the installation and ongoing running costs of the BWN. From a computational point of view, the gateway placement problem is a constrained combinatorial optimization problem. In this paper, we transform the gateway problem into a graph clustering problem and design a repairing genetic algorithm (RGA) to solve the graph clustering problem. Different from traditional GAs, this RGA embeds a procedure that can detect and repair those infeasible solutions generated by the crossover and mutation operators. Experimental results show that the infeasible solution detecting and repairing procedure can not only reduce the computation time of the RGA, but also improve the quality of the solutions generated by the RGA. In this paper, we also conduct an empirical study of the computational efficiency of the RGA. The analysis result shows that its computational efficiency is quadratic, which is computationally efficient.