Solutions for multiobjective optimization problems find use in architectures that support parallel processing. A new method proposed by Acan and Lotfi is seen to dominate over most other optimization solutions, as is the case with a typical Pareto-optimal set. The minimum requirement for dominance is that the dominant solution should not be worse than a competing solution in all objectives and should be better than the competitor for at least one objective.

The new approach proposed here converges to an optimal solution at a faster rate, thereby allowing the authors to claim it to be a superior architecture over most other algorithms.

This paper elaborates in detail on the literature on the existing state of the art for common multiobjective optimization approaches. The algorithms and approaches covered in this section include multiagent systems for single-objective optimization as well as for multiobjective optimization. Historical frameworks have been illustrated in detail with the help of block diagrams.

The method deployed here involves the use of multiobjective metaheuristic agents that work iteratively over a population of solutions in two phases. In the first phase, subpopulations of solutions are created based on dominance ranks of its elements, and in the second phase each multiobjective metaheuristic agent is assigned to work on its assigned subpopulation. The metaheuristic agents finally cooperate with each other, sharing their solutions.

Finally, the effectiveness of the new architecture is tested using the CEC2009 benchmark, which is a well-known set of benchmark problems. The results reflect the new approach of using a multiagent system of metaheuristic agents for solving multiobjective optimization problems possessing an advantage over other solutions.