A new improved fruit fly optimization algorithm for traveling salesman problem

Abstract

Traveling salesman problem (TSP) which is a classic combinational optimization problem has a wide range of applications in many areas. Many researchers focus on this problem and propose several algorithms. However, it was proved to be NP-hard, which is very difficult to be solved. No algorithm can solve any types of this problem effectively. In order to propose an effective algorithm for TSP, this paper improves the fruit fly optimization algorithm (FOA) proposed recently. As far as we know, the FOA has not yet been applied to solve TSP. Therefore, several modifications of FOA have to be made to meet the characteristics of TSP. Based on the whole search framework and the essence of FOA, some operations of particle swarm optimization (PSO) have been introduced into this method. In the smell search phase, the cluster mechanism of the fruit flies has been used to copy flies to one point and the mutation operation of genetic algorithm is used as the method of information exchanging among fruit flies for random search. In the visual search phase, the generalized PSO is applied to balance the global search and local search abilities of proposed algorithm. To evaluate the performance of proposed algorithm, some experiments and comparisons with other reported algorithms have been conducted. The results show the feasibility and effectiveness of proposed algorithm in solving TSP.