Multi-population Discrete Bat Algorithm with Crossover to Solve TSP

Abstract

Many meta-heuristic algorithms were proposed to solve several optimization problems. A new meta-heuristic bat algorithm (BA), inspired by the echolocation characteristics of micro-bats, has been extensively applied to solve continuous optimization problems. In addition, BA was also adapted to address combinatorial optimization problems. Unfortunately, like its basic version and other meta-heuristic algorithms, the adapted BA still suffers from some drawbacks such as slow speed convergence and easily trapping in local optima. We proposed a new variant of BA, called multi-population discrete bat algorithm (MPDBA), to solve traveling salesman problem (TSP). The validity of MPDBA was verified by comparative experiments using twenty TSP benchmark instances from TSBLIB. The experiments carried out show that MPDBA outperformed other state-of-art algorithms with respect to average and best solutions.