CLMGO: an enhanced Moss Growth Optimizer for complex engineering problems

• A novel metaheuristic algorithm, CLMGO, is proposed by integrating comprehensive learning into Moss Growth Optimization. • The Cryptobiosis mechanism is removed to enhance population stability and avoid premature convergence. • CLMGO demonstrates superior global exploration and local exploitation capabilities. • Extensive benchmarking on CEC 2017 functions shows CLMGO outperforms 13 state-of-the-art algorithms. • CLMGO is successfully applied to UAV path planning in eight scenarios, proving its practical effectiveness. While the bio-inspired Moss Growth Optimization (MGO) algorithm shows promise, it is hampered by critical deficiencies such as slow convergence, an imbalanced exploration-exploitation mechanism, and premature loss of population diversity. To overcome these limitations, this study proposes a Comprehensive Learning-enhanced Moss Growth Optimization (CLMGO) algorithm. The proposed CLMGO integrates a Comprehensive Learning (CL) strategy to enhance global search capabilities by enabling individuals to learn from multiple high-quality solutions. Concurrently, it streamlines the search process and improves computational efficiency by removing the redundant cryptobiosis mechanism. The superiority of CLMGO was rigorously validated through extensive experiments. On the CEC 2017 benchmark suite, CLMGO significantly outperformed the original MGO, seven classical metaheuristics, and six state-of-the-art algorithms, achieving top ranks in Friedman tests. Furthermore, CLMGO demonstrated remarkable effectiveness in real-world applications. In unmanned aerial vehicle (UAV) path planning, it achieved the optimal path cost in seven out of eight complex scenarios, and in these winning cases, it reduced the average path cost by 8.2% compared to the average of its competitors. Its binary form, bCLMGO, also dominated on 18 public feature selection datasets, where it ranked first for both classification accuracy and feature reduction. Notably, it achieved zero mean error on 16 datasets and reduced the average number of selected features by over 45% compared to the original bMGO. These comprehensive results establish CLMGO as a highly effective, robust, and versatile optimization tool, offering a reliable solution for complex scientific and engineering challenges.