A multi-swarm bat algorithm for global optimization

By simulating the echolocation behavior of bats in nature, bat algorithm (BA) is proposed for global optimization that is a recently developed nature-inspired algorithm. Since then, it has been widely used in various fields. Bat algorithm balance the global search and local search by adjusting loudness and pulse rate. However, there is so many loudness and pulse rate combinations that it is hard to choose the most proper one for different problems. In this paper, a multi-swarm algorithm, called multi-swarm bat algorithm (MBA), is proposed for global search problem. In MBA method, immigration operator is used to exchange information between different swarms with different parameter settings. Thus, this configuration can make a good trade-off between global and local search. In addition, the best individuals of every swarm is put into the elite swarm through selection operator. The bat individuals in elite swarm pass over next generation without performing any operators, and this can ensure these best solutions cannot be damaged during optimization process. In order to evaluate the efficiency of MBA method, MBA has been benchmarked by sixteen standard test functions by comparing with basic BA. The results show that the MBA method is able to search more satisfactory function values on most benchmark problems than BA.