An Improved NSGA-II Algorithm Based on Crowding Distance Elimination Strategy

Aiming at the diversity of Nondominated Sorting Genetic Algorithm II (NSGA-II) in screening out nondominated solutions, a crowding distance elimination (CDE) method is proposed. Firstly, the crowding distance is calculated in the same level of non-dominated solutions, and the solution of minimum crowding distance is eliminated; secondly, the crowding distance of residual solutions is recalculated, and the solution of minimum crowding distance is also eliminated. Repeat the above process, and stop the cycle when the nondominated solutions reaches the set number. In order to verify the effectiveness of the algorithm, experiments are carried out with the representative test functions: ZDT1, ZDT2, and ZDT3. The comparative experiments of NSGA-II, σ-Multi-objective particle swarm optimization algorithm (MOPSO), Non dominated sorting particle swarm optimization algorithm (NSPSO), and CDE were carried out respectively. By analyzing the diversity and convergence of the four algorithms, the strategy of nondominant solutions selection based on CDE has better performance.