Evolutionary harmony search algorithm with Metropolis acceptance criterion for travelling salesman problem

摘要

Harmony search (HS) algorithm is a new constructive meta-heuristic. In general, the intensification ability of a constructive meta-heuristic is not as good as that of iterative meta-heuristic, such as simulated annealing (SA) algorithm. To address this issue, we present a novel evolutionary HS (EHS) algorithm; in particular, we exploit the local search ability of SA algorithm to solve Travelling Salesman Problem (TSP). In EHS, we combine the evolution idea from evolutionary computation (EC) and the Metropolis acceptance criterion of SA algorithm to improvise a new harmony. EHS algorithm can achieve significantly better intensification ability by taking advantage of the evolution process of EC and the local search ability from SA. Furthermore, the probabilistic accepting criterion of SA can effectively keep EHS from premature stagnation. Simulation experiments of EHS were conducted based on benchmark TSP problems, and the results show that EHS algorithm has demonstrated promising performance in terms of solution accuracy and CPU time.