The Rapidly-exploring Random Tree (RRT) algorithm based on random sampling has been widely applied in global path planning for robots due to its collision-free and asymptotically optimal solution capabilities. However, in narrow and dynamic indoor environments, RRT and its optimized algorithms suffer from drawbacks such as multiple iterations and long planning times. While the Artificial Potential Field (APF) method can reduce the number of iterations and make the planned path smoother, it still faces issues such as local optima or inability to reach the target. To address these issues, this paper proposes the APF-IBRRT* algorithm, which combines the advantages of APF and B-RRT*. The algorithm first overcomes the local optimum problem of APF through the optimization of potential field factors and the introduction of virtual obstacles. Then, it designs a target threshold and proposes an adaptive step size search iteration strategy to address the issue of ineffective tree searches. Finally, the APF-IBRRT* algorithm is compared with B-RRT*, VPF-RRT*, and the classic A* algorithm. Simulation and actual experimental results demonstrate that the APF-IBRRT* algorithm exhibits superior performance in terms of planning time and iteration efficiency.

With the development of deep learning and sensors and sensor collection methods, computer vision inspection technology has developed rapidly. The deep-learning-based classification algorithm requires the acquisition of a model with superior generalization capabilities through the utilization of a substantial quantity of training samples. However, due to issues such as privacy, annotation costs, and sensor-captured images, how to make full use of limited samples has become a major challenge for practical training and deployment. Furthermore, when simulating models and transferring them to actual image scenarios, discrepancies often arise between the common training sets and the target domain (domain offset). Currently, meta-learning offers a promising solution for few-shot learning problems. However, the quantity of supporting set data on the target domain remains limited, leading to limited cross-domain learning effectiveness. To address this challenge, we have developed a self-distillation and mixing (SDM) method utilizing a Teacher–Student framework. This method effectively transfers knowledge from the source domain to the target domain by applying self-distillation techniques and mixed data augmentation, learning better image representations from relatively abundant datasets, and achieving fine-tuning in the target domain. In comparison with nine classical models, the experimental results demonstrate that the SDM method excels in terms of training time and accuracy. Furthermore, SDM effectively transfers knowledge from the source domain to the target domain, even with a limited number of target domain samples.

Weld defects in pressure vessels, prone to development under long-term operation under complex and harsh environments with alternating loads and corrosion, pose critical safety risks. Current defect classification primarily relies on manual interpretation of sensor images, which is inherently subjective. Therefore, deep learning (DL) has become a common way to classify images. However, insufficient quantity of defect sensor data limiting DL applicability, and the prevalence of small-scale, morphologically diverse defect instances that hinder accurate classification. To address these issues, we propose a two-stage meta-fine-tuning paradigm (TMF) for transfer learning in the classification of welding defects in pressure vessels. This paradigm migrates the weights of the source domain training to the RIAWELC dataset. The first stage employs meta-learning, where a limited representative sample is selected from multiple defect categories, allowing for rapid adaptation to new tasks with minimal supervision. The second phase performs fine-tuning with a small set of labeled target domain data and augments with MixUp to enhance generalization and robustness. Experimental results demonstrate that the TMF algorithm achieves an accuracy of over 92%, outperforming conventional manual inspection and DL approaches. The proposed paradigm demonstrates promising generalizability for defect classification tasks in analogous industrial scenarios.

Laser soldering has been gradually applied to the soldering of electronic components due to the rapid development of microelectronics. However, it is inefficient to use a mechanical shaft to move a laser beam. Here, a laser soldering system is constructed using galvanometer scanning, and an intelligent algorithm is also introduced to optimize the soldering path. Firstly, a laser soldering system for scanning of galvanometers is established, and the functions of visual monitoring, motion planning and parameter integration are presented. Secondly, the position of the laser beam and the corresponding soldering spot are determined, and the coordinate information is provided to plan a route by camera calibration and coordinate system transformation. Finally, the problem of path planning in this system is decomposed into the generation of the soldering point full coverage processing frame, and the route optimization of processing platform and laser beam motion. Furthermore, an improved clustering algorithm, based on the characteristics of system structure, and a hybrid optimization algorithm are designed to deal with the generation of the soldering point full coverage processing frame, the route optimization of processing platform and laser beam motion. In addition, the simulations and experiments are verified by test board. These findings shown that the established system and designed optimization algorithm can promote the efficiency of laser soldering.