Novel Consensus-Based Particle Swarm Optimization-Guided Surrogate-Enhanced Methodology for Solving Expensive Black-Box Optimization

A novel Consensus-Based Particle Swarm Optimization-Guided Surrogate-Enhanced Methodology (ACP-S) is developed to solve expensive black-box optimization (EBBO) problems. ACP-S formed form Global Exploration and Grouping Stage, Local Exploitation via the Surrogate Model Stage and Ranking, Refined Exploitation and Feedback Stage only search in a subset of the entire search space that contains high-quality optimal solutions and may include the global optimal solution. The proposed three-stage method is fast and deterministic in computing high-quality optimal solutions. Extensive experimental results demonstrate that this proposed method can obtain high-quality optimal solutions and outperforms several existing methods on small or large EBBO test functions.