Traveling Salesman Problem Solving Using Evolutionary Algorithms Guided by Complex Networks

Abstract

The evolutionary computation is an optimization strategy of the artificial intelligence, it is widely used for solving problems, which traditional algorithms cannot solve. Moreover, complex networks model systems, in which, the relationship structures are represented
as graphs. The Traveling Salesman Problem (TSP) is a well known computational problem solved using many strategies, included evolutionary computation. The literature review shows that the population of evolutionary computation strategies can be represented as complex networks, as follows: a graph represents the group of solutions, where the vertices (nodes) represent the individuals within the population of solutions, and the edges (links) represent the crosses among individuals (solutions). Thus, to improve solutions, links among existing nodes of the population are created. This work aims at the development of an algorithm to guide the population dynamics of an evolutionary algorithm that solves TSP. It is based on the hypothesis that one strategy guided by a complex dynamic network yields better results than a traditional algorithm. Furthermore, the results substantiate that the model proposed not only improves convergence compared to the traditional solution, but also enables shorter execution times. Strategies based on small-world networks show enhanced overall performance in the experiments conducted compared to
those using other network topologies, and subsequently the traditional solution.