MSE-CDO: A Multi-Strategy Enhanced Cloud Drift Optimizer for Global and Constrained Engineering Optimization

Abstract

Cloud Drift Optimization (CDO) is a recent nature-inspired metaheuristic with a simple adaptive search framework. Despite its adaptive mechanisms, CDO does not explicitly regulate initial space coverage, population dispersion during evolution, or agent-specific local refinement. This study proposes a Multi-Strategy Enhanced Cloud Drift Optimizer, termed MSE-CDO. First, randomized multistart maximin Latin hypercube sampling is employed to improve the spatial distribution of the initial population. Second, adaptive Gaussian mutation is introduced after the original CDO update to generate controlled perturbations according to the population dispersion and search stage, while greedy selection retains better candidates. Third, a directional pattern search mechanism is incorporated to strengthen local refinement through individual search directions and adaptive step lengths. The proposed algorithm is evaluated on unimodal and multimodal benchmark functions and constrained engineering design problems and is compared with the original CDO and several established metaheuristic algorithms. The experimental results show that MSE-CDO improves the performance of the original CDO on most benchmark functions, achieves the best overall Friedman rank among the compared algorithms, and obtains competitive solutions for the considered engineering problems. These results indicate that the proposed strategies improve the search and convergence performance of CDO while preserving its original framework.