Abstract: A multi-population co-evolutionary Memetic algorithm(MCMA) for symmetric TSP(traveling salesman problem) is presented.MCMA is devised on the basis of Memetic algorithm(MA) with 3 sub-populations co-evolution,and avoids premature convergence caused by poor population diversity of MAs.In MCMA,three different neighborhood search structures,including 2-exchange,3-exchange and PCV,are introduced for three sub-populations.The strategy effectively preserves the population diversity and converges quickly.Some instances in TSPLIB are shown to demonstrate the capabilities of the proposed algorithm,and some comparisons with other algorithms are made,such as the SGA(standard genetic algorithm),SMAs(simple Memetic algorithms) and GGA(greedy genetic algorithm).Simulation results of the algorithm are satisfactory,which proves the effectiveness of the proposed algorithm.