Locating facilities with busy servers using a genetic algorithm with simulation

Many systems rely on their ability to rapidly deliver services to customers. These systems depend upon the optimal allocation of servers within their network to achieve this goal. Determining optimal allocations is extremely complicated when the possibility of server unavailability exists. In this paper we analyze current techniques for solving our location problem and highlight some of the problems associated with these methods. We propose a genetic algorithm to locate facilities and servers to minimize the number of locations which cannot be covered within a specified time. Our proposed genetic algorithm is driven by the evaluation of the fitness of chromosomes through a discrete event simulation model. This method of evaluating chromosomes is used because of the difficulty of evaluating required probabilities. We apply our technique to an EMS system in Fayetteville, NC. We believe that our approach will lead to a general cost efficient technique for applicable location problems.