Many-Objective Particle Swarm Optimization Using Two-Stage Strategy and Parallel Cell Coordinate System.

Area of Science:

• Computational Intelligence
• Evolutionary Computation
• Multi-objective Optimization

Background:

• Balancing convergence and diversity is a key challenge in many-objective optimization evolutionary algorithms.
• Existing algorithms struggle to achieve both simultaneously, especially with a large number of objectives.

Purpose of the Study:

• To propose a novel algorithm, many-objective particle swarm optimization with the two-stage strategy and parallel cell coordinate system (PCCS), to enhance performance in many-objective optimization.
• To improve the comprehensive performance regarding convergence and diversity of the approximate Pareto front.

Main Methods:

• A two-stage strategy emphasizing convergence (single-objective optimizer) and diversity (many-objective optimizer) separately.
• A parallel cell coordinate system (PCCS) for managing diversity, maintaining archives, identifying dominance-resistant solutions, and selecting diversified solutions.
• Utilizing a leader group to select global best solutions, balancing population exploitation and exploration.

Main Results:

• The proposed PCCS algorithm demonstrated superior performance compared to six state-of-the-art algorithms.
• Performance was evaluated using inverted generational distance and hypervolume metrics on the DTLZ test suite.
• The algorithm effectively balances convergence and diversity, leading to better Pareto front approximation.

Conclusions:

• The proposed many-objective particle swarm optimization with PCCS effectively addresses the convergence-diversity trade-off.
• The two-stage strategy and PCCS are crucial components for achieving improved performance in many-objective optimization.
• The algorithm shows significant potential for solving complex many-objective optimization problems.