Abstract Surrogate-assisted intelligent algorithms (SAIAs) are effective for solving expensive multi-objective optimization problems, yet their performance often depends on labor-intensive manual design and high computational costs. To address these challenges, this paper proposes a hybrid-learning-driven automated surrogate-assisted multi-objective particle swarm optimization algorithm, termed HLASMPSO. The proposed framework integrates reinforcement learning-based optimizer selection, performance-guided surrogate adaptation, and evolutionary feedback-driven model management into a unified control architecture, enabling fully automated and adaptive optimization. By synergistically combining heterogeneous learning mechanisms, HLASMPSO substantially reduces human intervention and computational burden while improving robustness, generalization, and convergence performance. Extensive experiments on challenging benchmark functions and real-world municipal sewage treatment process (MSTP) optimization problems demonstrate that HLASMPSO consistently outperforms state-of-the-art algorithms.
Dai et al. (Thu,) studied this question.