Multi-objective particle swarm algorithm based on angular segmentation archive and dynamic update tactics.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Balancing solution convergence and diversity is a key challenge in multi-objective particle swarm optimization (MOPSO).
• Existing MOPSO algorithms often struggle to maintain solution quality and spread effectively across complex objective spaces.

Purpose of the Study:

• To propose a novel MOPSO algorithm, ASDMOPSO, designed to improve optimization efficiency.
• To enhance the management and diversity of non-dominated solutions through angular archiving and dynamic updates.

Main Methods:

• ASDMOPSO utilizes angular division of the external archive for solution management and diversity maintenance.
• A multi-stage initialization employs genetic and differential evolution algorithms for improved initial population quality.
• A dynamic flight parameter adjustment technique balances exploration and exploitation in real-time.

Main Results:

• ASDMOPSO demonstrated superior performance compared to several representative MOPSO algorithms across 22 benchmark functions.
• The algorithm achieved a significantly improved average IGD value of 0.032 on the ZDT4 test function.
• Statistical tests, sensitivity analysis, and complexity analysis validated the algorithm's effectiveness and efficiency.

Conclusions:

• ASDMOPSO offers a competitive and efficient approach to complex multi-objective optimization problems.
• The proposed angular archiving and dynamic parameter adjustment strategies effectively enhance solution convergence and diversity.
• The multi-stage initialization further contributes to the algorithm's overall performance.