The Crested Porcupine Optimizer (CPO), an emerging intelligent optimization algorithm, exhibits considerable potential for addressing complex engineering problems, yet its capabilities remain insufficiently investigated. Nevertheless, the original CPO is susceptible to premature convergence and suffers from insufficient population diversity. To effectively address these limitations, this paper proposes a multi-mechanism enhanced Crested Porcupine Optimizer (SDHCPO). Its core innovation lies in the integration of four key strategies: a Sobol-Opposition-Based Learning (Sobol-OBL) initialization strategy, which combines the Sobol sequence with opposition-based learning to generate an initial population that is more uniformly distributed in the high-dimensional search space; a cosine-annealing-based dynamic adjustment strategy that replaces the original random weights and substantially enhances convergence stability; the incorporation of the DE/rand/1 strategy in the first defense phase to disrupt positional dependence and prevent premature convergence; and a horizontal-vertical crossover strategy employed in the second defense phase to eliminate dimensional stagnation. Experimental results on two authoritative benchmark suites, CEC2017 and CEC2022, demonstrate that the proposed algorithm outperforms seven representative metaheuristic algorithms in terms of global exploration capability, local exploitation accuracy, and convergence robustness. Furthermore, empirical studies on five representative engineering design optimization problems show that SDHCPO consistently attains either the best-known solutions or highly competitive results reported in the literature, thereby further confirming its effectiveness and broad application potential for complex real-world engineering optimization tasks.
Xie et al. (Mon,) studied this question.