An Improved NSGAII Algorithm Based on Site-Directed Mutagenesis Method for Multi-Objective Optimization

Evolutionary algorithms have been greatly uti-lized in Multi-objective optimization problems. Existing studies on multi-objective evolutional algorithms (MOEAs) rarely consider the evolutionary environment which may influence the selection and mutation of the individuals in the evolutionary process. Therefore, with consideration of the evolutionary environment of human intervention, this paper proposes a novel site-directed mutagenesis method for MOEA to generate offspring based on reinforcement learning. In the evolutionary process, reinforcement learning is utilized to simulate the site-directed mutagenesis of human intervention, where key genes affecting the current status are identified through Q-learning agent in the mutation stage. The mutation stage is further applied in NSGAII and the new algorithms are named as RL-NSGAII. Different benchmark problems are used to verify the performance of the proposed algorithms through a large number of experiments. Compared with NSGAII, RL-NSGAII has advantages in the convergence speed of Pareto front and has outstanding performance in the diversity and stability of the solution set for both the two-objective and multi-objective benchmark problems.