A Novel Pareto Archive Evolution Algorithm with Adaptive Grid Strategy for Multi-objective Optimization Problem

Multi-objective evolutionary algorithms usually utilize fixed evolutionary mechanism and the evolutionary operators are static during the process of algorithm evolution. It is easy to cause a simple population structure, unable to exploit the search space fully and trapped in local optimal solution. In this paper, a novel method named Pareto Archive Evolution Strategy (PAES) with adaptive grid strategy (AGS_PAES) which only makes one mutation to create one new solution and use an “archive” which are called Non-Dominated Archive to store the best solution, is introduced. This procedure is completed by a special approach - adaptive grid method, which decides the criterion of the solution to be archived and the place of the grid location the solution would be stored. The Pareto front obtained by the procedure outperforms the classical Multi-objective Genetic Algorithm (MOGA). Simulation results on the standard benchmark problems show that the proposed adaptive scheme has a better convergence and diversity compared with the second generation classical multi-objective evolutionary algorithms.