Previous research on multiple traveling salesman problem were mostly limited to the kind that employed total-path-shortest as the evaluating rule, but little notice was paid attention to the kind that employed longest-path-shortest as the evaluating rule. In order to solve this problem, an improved differential evolution algorithm was proposed to optimize it. In these methods, real number
... [Show full abstract] encoding and roulette wheel selection were adopted for improved differential evolution and neighborhood search operator was added to it. It was fit for solving symmetric and asymmetric multiple traveling salesman problem. Asymmetric multiple traveling salesman problems were simulated. Through comparison of results, it is shown that improved differential evolution algorithm is efficient to solve the kind of discrete combinatorial problem, such as multiple traveling salesman problems.