Robust Multiobjective Competitive Swarm Optimization Based on Evolutionary Trend Prediction.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Swarm intelligence

Background:

• The competitive swarm optimizer (CSO) is effective for multiobjective optimization but suffers from evolutionary process uncertainty, impacting reliability.
• Existing methods struggle to mitigate the inherent randomness in CSO's evolutionary path.

Purpose of the Study:

• To propose a robust multiobjective CSO with a predictive indicator (RMOCSO-PI) to enhance algorithmic reliability and reduce inefficient searches.
• To improve the stability and performance of multiobjective optimization using a predictive approach.

Main Methods:

• Developed a predictive indicator using an autoregressive model to forecast evolutionary trends based on historical swarm data.
• Classified particles into winners and losers for differential evolution guidance, employing a space fusion-based competitive mechanism for precise direction.
• Introduced a dynamic cooperative mechanism with three patterns for purposeful diversity exploration based on estimated evolutionary states.

Main Results:

• The RMOCSO-PI approach effectively reduces aimless and inefficient searches, thereby enhancing algorithmic robustness.
• Convergence analysis provides theoretical support for the validity and reliability of the proposed RMOCSO-PI.
• Experimental results demonstrate that RMOCSO-PI achieves more stable and excellent optimization performance compared to existing methods.

Conclusions:

• RMOCSO-PI significantly improves the robustness and reliability of multiobjective optimization by addressing evolutionary uncertainty.
• The integration of predictive indicators, differential guidance, and dynamic cooperation leads to superior and stable optimization outcomes.