Estratégias bio-inspiradas aplicadas em problemas discretos com muitos objetivos

Abstract

Multi-objective optimization problems are very common in the day-to-day life and come up in many fields of knowledge. In this work, several strategies for multi-objective optimization have been explored and compared on two well known discrete problems in computer science: the knapsack problem and the multicast routing problem. Among all strategies to solve multi-objective discrete problems, genetic algorithms (GAs) and ant colony optimization (ACO) are the ones which generally provide the best results. In this work both approaches are explored through several experiments involving 10 different algorithms. All of the algorithms evaluated were adapted to the proposed problems. Furthermore, a new algorithm has been proposed, the Many-objective Ant Colony Optimization based on Non-Dominated Sets (MACO/NDS), which has been evaluated and compared against all other methods investigated here, considering different performance metrics. In many scenarios, it has been capable of finding superior sets of solutions and took less time to execute.