GACO – A HYBRID ANT COLONY OPTIMIZATION METAHEURISTIC FOR THE DYNAMIC LOAD-BALANCED CLUSTERING PROBLEM IN AD HOC NETWORKS

Abstract

We present the design of a novel hybrid genetic ant colony optimization (GACO) metaheuristic. Genetic ant colony optimization is designed to address the dynamic load-balanced clustering problem formulated from ad hoc networks. The main algorithm in GACO is ACO. Whenever an environment change occurs (i.e., nodes move), it makes use of a genetic algorithm to quickly achieve adaptation by refocusing the search process around promising areas of the search space induced by the new problem structure. Compared to other ACO approaches for dynamic problems, GACO does not depend on any problem-specific heuristics to repair or deconstruct solutions. Genetic ant colony optimization also does not require the knowledge of the specific changes that occurred. We compare GACO with three other adaptation methods, namely, P-ACO, PAdapt, and GreedyAnts. P-ACO is a population-based ACO approach that invokes a repair algorithm on its population of solutions when an environment change occurs. PAdapt works by adapting the values of major ACO parameters, while GreedyAnts employs a group of ants that construct solutions in a greedy manner. Empirical results show that GACO is able to react and recover faster from any performance degradation. Genetic ant colony optimization also produces better solutions within the allowable recovery window. These results are shown to be statistically significant.