The Research On Non-uniform Problems Of Multi-objective Evolutionary Algorithm

Multi-objective evolutionary algorithm has become an important research area in recent years, and many algorithms have been proposed. Researchers believe that true efficiency of an algorithm reveals when it is applied to difficult and challenging test problems, and not to easy problems. In order to seek problem features causing difficulty to an algorithm, a lot of difficult problems are constructed: Multi-modal multi-objective problem, Deceptive multi-objective optimization problem, and the problem where the Pareto-optimal front is non-uniform or discontinue, etc. There exist some multi-objective optimize problems in which the Pareto-optimal front is non-uniform or discontinue. In these problems, the existed evolutionary algorithms cannot correctly reflect the distribution of true solutions. In this paper, a grids technique based on niching evolutionary algorithm for multi-objective optimization (GNEA) is proposed, which aims to solve non-uniform problems. GNEA applies niching technique to keep the non-uniform distribution of local solutions, and grid division to maintain diversity of the globe solutions. Compared with the existed algorithms, the experimental results demonstrate that the GNEA could obtain a well distribution of the solutions that corresponds to the true distribution on non-uniform multi-objective optimization problems. In addition, the time consumption of GNEA is low. It is clear that GNEA is a kind of efficient multi-objective optimization algorithm, which can be adapted to non-uniform problems.