This paper presents an enhanced artificial lemming algorithm (EALA) for solving complex unmanned aircraft system (UAV) path planning problems in three-dimensional environments. Key improvements include chaotic initialization, adaptive perturbation, and hybrid mutation, enabling a better exploration–exploitation balance and local refinement. Validation on the IEEE CEC2017 and CEC2022 benchmark functions demonstrates the EALA’s superior performance, achieving faster convergence and better algorithm performance compared to the standard ALA and 10 other algorithms. When applied to UAV path planning in large- and medium-scale environments with realistic obstacle constraints, the EALA generates Pareto-optimal paths that minimize length, curvature, and computation time while guaranteeing collision avoidance. Benchmark tests and realistic simulations show that the EALA outperforms 10 algorithms. This method is particularly suited for mission-critical applications with strict safety and time constraints.

The starfish optimization algorithm (SFOA) is a metaheuristic evolutionary intelligence algorithm with a great global search capability and strong adaptability. Although the SFOA has a good global search capability, it is not accurate enough in local search and converges slowly. To further enhance this convergence ability and global optimization ability, an enhanced starfish optimization algorithm (SFOAL) is proposed that combines sine chaotic mapping, t-distribution mutation, and logarithmic spiral reverse learning. The SFOAL can remarkably enhance both the global and local convergence capabilities of the algorithm, leading to a more rapid convergence speed and greater stability. In total, 23 benchmark functions and CEC2021 were used to test the development, search, and convergence capabilities of the SFOAL. The SFOAL was compared in detail with other algorithms. The experimental results demonstrated that the overall performance of the SFOAL was better than that of other algorithms, and the joint strategy could effectively balance the development and search capabilities to obtain stronger global and local optimization capabilities. For solving practical problems, the SFOAL was used to optimize the back propagation (BP) neural network to solve the ultra-wideband line-of-sight positioning problem. The results showed that the SFOAL-BP neural network had a smaller average position error compared to the random BP neural network and the SFOA-BP neural network, so it can be used to solve practical application problems.

The Walrus Optimization (WO) algorithm, a metaheuristic inspired by walrus behavior, is known for its competitive convergence speed and effectiveness in solving high-dimensional and practical engineering optimization problems. However, it suffers from a tendency to converge to local optima and exhibits instability during the iterative process. To overcome these limitations, this study proposes an improved WO (IMWO) algorithm based on the integration of Differential Evolution/best/1 (DE/best/1) mutation, Logistics–Sine–Cosine (LSC) Mapping, and the Beta Opposition-Based Learning (Beta-OBL) strategy. These strategies work synergistically to enhance the algorithm’s global exploration capability, improve its search stability, and accelerate convergence with higher precision. The performance of the IMWO algorithm was comprehensively evaluated using the CEC2017 and CEC2022 benchmark test suites, where it was compared against the original WO algorithm and six other state-of-the-art metaheuristics. Experimental data revealed that the IMWO algorithm achieved average fitness rankings of 1.66 and 1.33 in the two test suites, ranking first among all compared algorithms. The WSN coverage optimization problem aims to maximize the monitored area while reducing perception blind spots under limited node resources and energy constraints, which is a typical complex optimization problem with multiple constraints. In a practical application addressing the coverage optimization problem in Wireless Sensor Networks (WSNs), the IMWO algorithm attained average coverage rates of 95.86% and 96.48% in two sets of coverage experiments, outperforming both the original WO and other compared algorithms. These results confirm the practical utility and robustness of the IMWO algorithm in solving complex real-world engineering problems.

Adaptive Reinforced Gray Langur Optimization (ARGLO), an enhanced variant of the Gray Langurs Optimizer, is developed for high-dimensional, multimodal, and nonlinear search landscapes susceptible to local trapping. Although the original GLO performs multi-population cooperative search by simulating the social structures of gray langurs, it still suffers from uneven random initialization, insufficient adaptive population partitioning, weak local perturbation, and premature convergence. ARGLO incorporates three strategies: good point set-based oppositional and quasi-oppositional learning initialization, hierarchical equilibrium adaptive population partitioning, and elite-guided hybrid mutation. Collectively, these mechanisms generate a higher-quality starting population, coordinate global search with local refinement, and reduce the risk of entrapment in suboptimal regions. Evidence from component-wise experiments together with the CEC test suite indicates that ARGLO delivers higher solution precision, steadier convergence, as well as more consistent performance, especially as dimensionality increases. Moreover, ARGLO is applied to near-infrared spectral feature selection and SVR parameter optimization for polysaccharide content prediction in Dendrobium huoshanense. Compared with unoptimized SVR, ARGLO-SVR reduces RMSE by 35.35% and improves R² by 21.92%; compared with GLO-SVR, it reduces RMSE by 6.05% and improves R² by 2.30%. These results demonstrate the effectiveness and application potential of ARGLO in complex optimization and rapid nondestructive quality detection of traditional Chinese medicinal materials.

This paper proposes DyMC-YOLO, an efficient RGB–infrared object detector based on YOLOv13. The evaluated infrared modalities include thermal infrared in the M3FD benchmark and near-infrared in a self-constructed RGB–NIR grape dataset. To balance cross-modal interaction with computational cost, the model integrates a streamlined dual-stream backbone, Complementary Mask-Guided Feature Fusion (CMFF), and a Dynamic Multi-Path Detection Head (DMP-Detect). After validation-based checkpoint selection, final evaluation on held-out test sets yielded mAP@0.5:0.95 scores of 51.73% on M3FD and 83.06% on the grape dataset, outperforming EarlyFusion-YOLOv13 by 4.31 and 1.75 percentage points, respectively. Operating at 45.87 FPS with 8.026 M parameters and 19.216 GFLOPs on an NVIDIA L40S GPU, the model provides an accurate and computationally efficient solution for multimodal RGB–infrared object detection.