Research On Archive Maintenance Strategy In Multi-objective Evolutionary Algorithm

In real life, there are a large number of optimization problem, and the objective number more than 3 are called many-objective optimization problem, is greater than 1 and less than 3 called multi-objective optimization problem. Compared with the optimization problem which objective number less than or equal to 3,Many-objective optimization problem in search for the optimal solution more difficult, often faced with lack of selection pressure, hardly maintain diversity and visualization. Archive is the sign of the second generation of evolutionary algorithm,and was applied successfully in a large number of classical Multi-objective optimization algorithm, but under the Many-objective problem, Pareto-dominance ineffective make lots of archive maintenance strategies based on it ineffective too,and also lead the algorithm convergence to the true front difficult.Based on above problems, this paper carried out deep research on archive strategies. The study mainly includes two aspects: Firstly, try to balance search in Multi-objective optimization algorithm using archive maintenance strategy; Secondly,focus on the expand of archive maintenance strategy in Many-objective problems optimization to solve the problem of lack of selection pressure. Based on a big amount of reference, this paper carried out the following research work:1. In view of imbalance of global search and local search in multi-objective evolutionary algorithms, designed a new archive maintenance strategy, which effectively avoid the algorithm falls into local optimum. With the comparison of the same type hot algorithms, agMOPSO and pdMOPSO, on standard test functions,demonstrates that the convergence and diversity of balance search strategy on the Pareto front have significantly increased.2. In view of the shortage of convergence in many-objective evolutionary algorithms, proposed a new archive maintenance strategy DRS-PCCS based on dimension reduction, which has improved the selection pressure of evolutionary progress. The strategy DRS-PCCS which proposed in this paper has been applied to the NSGA-? in the NSGA-?-DP algorithm.Through experiment contrast with the original algorithm NSGA-?, proved that the new algorithm can improve the NSGA-? to the problem of shortage of convergence and can bring new development to solve the Many-objective optimization problem.Above work has digged deeply into the research value of archive under the Multi-objective and Many-objective optimization problem. Not only strength the global search of algorithm and achieve the goal of balancing search progress, but also design a new archive maintenance strategy to effectively promote the Many-objective optimization problem.