Compared to the traditional non-load-bearing multi-legged robots, the heavy-duty multi-legged robots typically not only have larger body weight, larger volume, and larger load ratio but also require greater energy dissipation. Traditional path planning often focuses on the problem of finding the shortest path. However, the substantial load capacity and multi-jointed structure of heavy-duty multi-legged robots impose stringent requirements on path smoothness. Consequently, the smoothness requirement makes the traditional A* algorithm unsuitable for applications where low-energy operation is critical. An improved low-energy path planning method based on the A* algorithm is presented for an electrically driven heavy-duty six-legged robot. Then, the environment is discretized by using the grid method to facilitate path searching. To address the path zigzagging problem caused by the traditional A* algorithm, the Bézier curve smoothing technique is adopted. The continuous curvature transitions are employed to significantly improve the smoothness of path. The heuristic function in the A* algorithm is enhanced through a dynamic weight adjustment mechanism. The nonlinear suppression strategy is introduced to prevent data changes and improve the robustness of the algorithm. The effectiveness of the proposed method is verified through the MATLAB simulation platform system. The simulation experiments show that, in various environments with different obstacle densities (0.17–0.37%), compared with the traditional A* algorithm, the method proposed in this paper reduces the average path length by 7.2%, the number of turning points by 25.9%, and the energy consumption by 5.75%. The proposed improved A* algorithm can significantly overcome the problem of insufficient smoothness in traditional A* algorithms and reduce the number of nodes generated by the control data stack, which improves the optimization efficiency during path planning. As a result, the heavy-duty six-legged robots can walk farther and operate for longer periods of time while carrying the limited energy sources.

Casting defect detection is a crucial link in industrial production, directly affecting product quality and safety. Traditional manual inspection methods, due to their low efficiency, strong subjectivity, and high missed detection rates, can no longer meet the demands of modern intelligent manufacturing. In recent years, the breakthrough progress of deep learning technology in the field of image recognition has provided new solutions for automated casting defect detection. This paper systematically reviews automatic casting defect recognition methods based on deep learning, focusing on aspects such as algorithm improvements, practical applications, and future development directions. It discusses in detail the optimization strategies of object detection algorithms (such as the YOLO series, Mask R-CNN) and semantic segmentation algorithms (such as U-Net, Retina Net), and summarizes the latest achievements in industrial inspection system development. Finally, it prospects future research directions including self-supervised learning, unsupervised learning, and real-time quantitative analysis, aiming to promote the development of casting inspection technology towards intelligence and automation.

Based on the lunar exploration project, a mechanical structure design scheme for a heavy-duty hexapod robot movement system with bounce characteristics was proposed. The materials used were determined by the robot's working environment and performance standards. The mechanical structure of a jumping heavy-duty hexapod robot was designed using SolidWorks 3D modeling software. Based on the analysis of the working modes of the root joint, hip joint, knee joint, and jumping energy storage mechanism, the main components of each joint and jumping energy storage mechanism were designed, and a single leg mechanism was obtained. Based on the special soil structure on the lunar surface, the foot end buffer mechanism was designed, and finite element stress analysis was conducted on the key components of the robot model, proving the rationality of the mechanical structure of a heavy-duty hexapod robot with bounce characteristics.