In this paper, an enhanced differential evolution algorithm based on dual population hybrid update and random population replacement strategy (namely RPR mechanism) is proposed, which is called DHRDE. DHRDE algorithm involves three key improvements, first, the elite reverse population is constructed according to the original population before the update phase to uncover more potential areas to be searched. Second, a perturbation mechanism is integrated into the DE/rand/2 approach of the differential evolution algorithm to bolster its search efficiency, two updating models are established using co-leadership of random and locally optimal individuals, and then dual-population hybrid update strategy is adopted to achieve all-round and multi-angle search. Thirdly, using RPR mechanism to operate multiple types of mutations on some populations further improves the convergence accuracy. In order to verify the effectiveness of the proposed algorithm, DHRDE is compared with a variety of different types of algorithms in multi-dimension of the CEC2017, CEC2020 and CEC2022 test set, and statistical analysis is performed by Wilcoxon rank sum test and Friedman test. The results show that DHRDE algorithm has better performance. DHRDE algorithm is also used to solve seven engineering design problems and three PV model parameter estimation problems, the optimization results show that DHRDE algorithm is suitable for different complex problems and has effectiveness. In addition, this paper establishes a smooth path planning model for multi-size robots, and uses DHRDE to solve the model, the results of five groups of simulation experiments show that DHRDE algorithm can provide robot moving trajectories with higher smoothness and shorter paths. Analyzing and comparing the fitness metrics through heat maps, the comparative study demonstrates that the DHRDE algorithm is more advantageous and stronger than other algorithms in solving the smooth path planning model for multi-size robots. The above results show that DHRDE algorithm has better performance and has great advantages and competitiveness in solving engineering application optimization problems.

中文翻译：

---

DHRDE：面向工程应用的基于双群体混合更新和 RPR 机制的差分进化算法


本文提出了一种基于双种群杂交更新和随机种群替换策略的增强差分进化算法（即 RPR 机制），称为 DHRDE。DHRDE 算法涉及三个关键改进，首先，在更新阶段之前根据原始种群构建精英反向种群，以发现更多潜在搜索区域。其次，在差分进化算法的 DE/rand/2 方法中集成扰动机制以提高其搜索效率，利用随机个体和局部最优个体的共同领导建立两个更新模型，然后采用双群体混合更新策略实现全方位、多角度搜索。再次，使用 RPR 机制对某些种群进行多种类型的突变操作，进一步提高了收敛精度。为了验证所提算法的有效性，将 DHRDE 与多种不同类型的算法在多维CEC2017、CEC2020和CEC2022测试集上进行比较，并通过 Wilcoxon 秩和检验和 Friedman 检验进行统计分析。结果表明，DHRDE 算法具有更好的性能。DHRDE 算法还用于解决 7 个工程设计问题和 3 个 PV 模型参数估计问题，优化结果表明 DHRDE 算法适用于不同的复杂问题，具有有效性。此外，该文建立了多尺寸机器人的平滑路径规划模型，并使用 DHRDE 对模型进行求解，五组仿真实验结果表明，DHRDE 算法能够为机器人运动轨迹提供更高的平滑度和更短的路径。 通过热图分析和比较适应度指标，比较研究表明，在求解多尺寸机器人的平滑路径规划模型方面，DHRDE 算法比其他算法更具优势和强度。以上结果表明，DHRDE 算法具有较好的性能，在解决工程应用优化问题方面具有很大的优势和竞争力。