Local Self-Adaptive Mutation: Enhancing Multi-Objective Optimization through Localized Operator Selection and Parameter Adaptation

This paper tackles the challenge of simultaneously selecting and tuning mutation operators in multi-objective evolutionary optimization. The proposed Local Self-Adaptive Mutation extends the Adaptive Non-dominated Tournament Genetic Algorithm. Unlike global approaches, it selects operators based on effectiveness in specific regions of the objective space. A key innovation is the adaptive mutation probability, replacing predefined values and allowing dynamic adjustment across the Pareto Front. Applied to two real-world problems—Multi-Skill Resource Constrained Project Scheduling and Multi-Stage Resource Allocation. The enhanced method matched or outperformed manually tuned approaches.