An empirical comparison of Tabu Search, Simulated Annealing, and Genetic Algorithms for facilities location problems

Algorithms to solve Facility Location Problems (FLP) optimally suffer from combinatorial explosion and resources required to solve such problems repeatedly as required in practical applications become prohibitive. In these cases heuristic methods are the only viable alternative. We compare the relative performance of Tabu Search (TS), Simulated Annealing (SA) and Genetic Algorithms (GA) on various types of FLP under time-limited, solution-limited, and unrestricted conditions. The results indicate that TS shows very good performance in most cases. The performance of SA and GA are more partial to problem type and the criterion used. Thus, in general we may conclude that TS should be tried first to the extent that it always yields as good or better results and is easy to develop and implement.