A parallel genetic algorithm to solve the set-covering problem

Abstract

This work presents a parallel genetic algorithm (PGA) model to solve the set-covering problem (SCP). Experimental results obtained with a binary representation of the SCP, show that—in terms of the number of generations (computational time) needed to achieve solutions of an acceptable quality—PGA performs better than the sequential model. This comportment can be explained principally because, the PGA of p nodes—each one with its corresponding local population P L—behaves like a sequential GA with a global population, P G, of the same size, which it—the sequential GA—has the great disadvantage of having to completely evaluate in each generation. Not so the PGA, which only evaluates a pth part of the P G. Scope and purpose Since classical optimization techniques are inefficient to solve NP-complete problems—in terms of computational complexity—new methods have been developed. The genetic algorithm (GA), is one of such methods proposed to solve combinatorial optimization. Although Genetic Algorithms (GAs) are efficient to solve these kinds of hard problems, during recent years, models of parallel genetic algorithms (PGAs) have been used to improve both quality of solutions and computing time. The aim of using PGAs, is to discover how the interchange of genetic information of separate populations, affects or influences the final solution. The exploration of different solution spaces could optimize the search in terms of both computational time and quality of solution.