Hybridizing Metaheuristics for Multi and Many-objective Problems in a Multi-agent Architecture
Metaheuristic hybridization. Multi-objective optimization. Agent intelligence.Multi-agent paradigm. Decomposition.
Hybrid algorithms combine the best features of individual metaheuristics. They haveproven to be effective in finding high-quality solutions for multi-objective combinatorialoptimization problems. Architectures provide generic functionalities and features forimplementing new hybrid algorithms capable of solving arbitrary optimization problems.Architectures based on agent intelligence and multi-agent concepts, such as learning andcooperation, give several benefits for hybridizing metaheuristics. Nevertheless, there isa lack of studies on architectures that fully explore these concepts for multi-objectivehybridization. This thesis studies a multi-agent architecture namedMO-MAHM, inspiredby Particle Swarm Optimization concepts. In theMO-MAHMarchitecture, particles areintelligent agents that learn from past experiences and move in the search space, lookingfor high-quality solutions. The main contribution of this work is to study theMO-MAHMpotential to hybridize metaheuristics for solving problems with two or more objectives.We investigate the benefits of machine learning methods for agents’ learning supportand propose a novel velocity operator for moving the agents in the search space. Theproposed velocity operator uses a path-relinking technique and decomposes the objectivespace without requiring aggregation functions. Another contribution of this thesis isan extensive survey of existing multi-objective path-relinking techniques. Due to a lackin the literature of effective multi- and many-objective path-relinking techniques, wepresent a novel decomposition-based one, referred to asMOPR/D. Experiments includethree differently structured combinatorial optimization problems with up to five objectivefunctions: 0/1 multidimensional knapsack, quadratic assignment, and spanning tree. Wecompared theMO-MAHMwith existing hybrid approaches, such as memetic algorithmsand hyper-heuristics. Statistical tests show that the architecture presents competitiveresults regarding the quality of the approximation sets and solution diversity