R2-Based Multi/Many-Objective Particle Swarm Optimization

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Multi-objective Optimization

Background:

• Multi-objective optimization problems (MOPs) present challenges in finding optimal solutions.
• Existing methods often rely on Pareto dominance or external archives, which can be complex.
• Handling many-objective optimization problems (MaOPs) requires efficient search guidance strategies.

Purpose of the Study:

• To propose a novel approach for multi/many-objective optimization by coupling the R2 performance measure with Particle Swarm Optimization (PSO).
• To demonstrate that this coupling can guide the search process effectively without using Pareto dominance or external archives.
• To validate the efficacy of the proposed method on various test problems and compare it with existing Multi-Objective Evolutionary Algorithms (MOEAs).

Main Methods:

• Integration of the R2 performance measure with Particle Swarm Optimization (PSO).
• Development of a well-designed interaction process to maintain the core metaheuristic.
• Validation using standard test problems and performance metrics from the literature.
• Comparative analysis against four well-known MOEAs and an indicator-based MOEA for many-objective problems.

Main Results:

• The proposed R2-PSO approach yields competitive results compared to four established MOEAs on multi-objective problems.
• The method demonstrates significant strength in many-objective optimization scenarios.
• It outperforms a well-known indicator-based MOEA specifically in many-objective problem settings.

Conclusions:

• The R2 performance measure, when coupled with PSO, offers an effective alternative for multi/many-objective optimization.
• This approach simplifies the search process by avoiding Pareto dominance and external archives.
• The method shows particular promise and superior performance in the challenging domain of many-objective optimization.