A Novel Hybrid Optimizer Based on Coati Optimization Algorithm and Differential Evolution for Global Optimization and Constrained Engineering Problems

Abstract This paper presents a novel hybrid metaheuristic, the Hybrid Coati Optimization Algorithm with Differential Evolution (HCOADE), developed to address complex global optimization tasks and constrained engineering design problems. HCOADE integrates the exploration-driven behaviour of the Coati Optimization Algorithm (COA)-inspired by the social foraging and predation strategies of coatis-with the powerful mutation and crossover mechanisms of Differential Evolution (DE), thereby achieving a balanced and adaptive search process. The hybridization enhances global exploration and local exploitation, enabling the algorithm to efficiently navigate diverse and challenging optimization landscapes. To rigorously evaluate its performance, HCOADE is tested on benchmark suites from CEC 2014, 2017, 2020, and 2022, which encompass unimodal, multimodal, hybrid, and composition functions. It is also applied to real-world constrained engineering problems, such as pressure vessel design, cantilever beam optimization, and reinforced concrete beam design. Comparative experiments against state-of-the-art algorithms—including COA, DE, RSA, PSO, SSA, BBO, QIO, DMOA, and others—demonstrate that HCOADE consistently delivers superior solution quality, faster convergence, and higher robustness. Quantitative results show that HCOADE achieved the 1st place average rank across all four benchmark suites. It obtained top performance on 80% of the functions in CEC 2014, 66.7% in CEC 2017, 70% in CEC 2020, and 66.7% in CEC 2022. Furthermore, HCOADE outperformed or matched CEC competition-winning algorithms such as LSHADE-cnEpSin, LSHADE-SPACMA, and CMA-ES on numerous challenging functions of CEC 2017. Statistical analyses, including Wilcoxon Rank Sum Tests and ranking evaluations, confirm the significance and reliability of HCOADE’s performance. Furthermore, convergence behaviour, measurement of exploration and exploitation, and sensitivity analysis highlight the algorithm’s adaptability and stability across varied problem domains. This study contributes a computationally efficient and generalizable hybrid optimization framework, offering a promising solution for theoretical benchmarks and real-world engineering applications.