Key points are not available for this paper at this time.
The construction of large bridges requires extensive prefabrication of steel components, where the efficiency and precision of welding operations directly influence construction progress and quality. In the automated production of steel components, the traditional manual teaching programming method for welding robotic arms is inefficient, and the lack of support for automatic path planning has impeded the development of intelligent autonomous operations. This study proposes a method for segmenting the work area for large-scale panel unit components with multiple transverse and longitudinal ribs, considering the operational range of gantry welding robots, to enable step-by-step welding. To address the need for controlling welding deformation in actual operations, the degree of constraint is introduced for planning the welding sequence. Sampling-based rapidly exploring random tree (RRT) algorithms are commonly used to plan collision-free trajectories for robotic arms in three-dimensional (3D) space. This study integrates improvements from the RRT* and RRT-connect algorithms into the basic RRT framework, incorporating adaptive sampling space and a sampling point selection mechanism. The enhancements improve path search efficiency by limiting the sampling space during each search. Additionally, the algorithm conducts multiple samples at each step, selecting the point closest to the vertical plane through the start and target points, effectively shortening the path length. Experimental results in complex scenarios demonstrate that the proposed improved RRT algorithm reduces path cost and time cost by approximately 30% and 70%, respectively, compared with RRT* and other enhanced algorithms, with more stable path planning outcomes.
Hou et al. (Thu,) studied this question.