Abstract: Proposed in this paper is a multi-strategy discrete harmony search ( MDHS) algorithm for solving traveling salesman problem ( TSP) .In the algorithm,a discrete improvisation is designed by combining a -opt algorithm,and three strategies are employed together to improve its global optimization ability.The teaching-learningbased optimization strategy is adopted to provide a new way of producing new harmonies,so as to improve the quality of harmony memory ( HM) .The elite perturbation strategy is constructed to explore the neighborhoods of the best harmony constantly to perform a fine local search,so that the convergence precision of the proposed algorithm can be increased.The sort-selection-based update strategy is designed to preserve the diversity of HM for the sake of avoiding premature convergence.Experimental results show that MDHS can solve TSP effectively,and it holds an excellent search performance no matter in convergence speed or precision.

Key words: harmony search algorithm, traveling salesman problem, -opt algorithm, discrete improvisation, multi-strategy