.png){kind=icon}
.png){kind=icon}
Return
A constraint violation-based adaptive switching surrogate-assisted evolutionary algorithm for expensive constrained multiobjective optimization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2026.102372.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
When solving expensive constrained multi-objective optimization problems (ECMOPs), the number of constraints is often large, which poses a significant challenge to existing surrogate-assisted evolutionary algorithms (SAEAs), especially when computational resources are very limited. Due to the high complexity introduced by the large number of constraints, the computational cost of building and applying surrogate models can become prohibitively expensive, thus affecting the solution quality. To address this issue, this paper proposes a constraint violation-based adaptive switching surrogate-assisted evolutionary algorithm, which adaptively builds and applies different surrogate models for different parts of the population. The algorithm first accelerates convergence through unconstrained search, and then adaptively divides the search process into three phases based on the constraint satisfaction status: the fully infeasible phase, the partially feasible phase, and the fully feasible phase, thus refining the model building and application. Additionally, the algorithm employs three types of archives with different update criteria to assist in building surrogate models and reduce the impact of errors introduced by surrogate models. Experimental results show that, on 36 benchmark problems and six real-world engineering problems, the proposed algorithm outperforms eight state-of-the-art competitors, highlighting its strong competitiveness in solving ECMOPs.
Keywords:
surrogate-assisted evolutionary algorithm
constrained multi-objective optimization
adaptive switching
constraint violation
expensive optimization
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W