Autonomous robotic welding path planning for small‑gap assembly joints in complex steel structures

In steel structure, closed and semi-closed joints are widely used because of their importance in corner and specialized connection assemblies. These joints exhibit a variety of shapes, which are typically formed according to design specifications. However, their assembly sequence is heavily constrained by spot-welding locations and gripping points, and the limited workspace often gives rise to interference, posing significant challenges to the assembly process. This paper proposes a novel approach that leverages digital twins, a vision-based recognition system, Monte Carlo Tree Search, and path-planning algorithms to generate optimal spot-welding and assembly sequences, thereby improving both efficiency and quality. The method effectively addresses the complexity and uncertainty inherent in angle-box assembly for steel structures, ultimately enabling the search for feasible solutions across arbitrarily shaped angle-box nodes.