Development Status and Trends of Welding Robot Technology for Steel Structures

Steel structure fabrication and erection are characterized by large component dimensions， a high degree of non-standardization， small-batch and high-variety production， pronounced assembly deviations and welding-induced thermal deformation， a high proportion of multi-layer and multi-pass welds in thick plates， and substantial differences between factory fabrication and on-site erection conditions. Consequently， teach-and-playback welding robots， which have been widely adopted in repetitive manufacturing industries such as the automotive industry， cannot be directly transferred to steel structure applications. To clarify the current technological stage， applicable scenarios， and future development directions of welding robots for steel structures， this paper reviews representative application scenarios， including steel bridge panel units， factory fabrication of building steel structures， complex thick-plate joints， and on-site erection， following the logical sequence of technological evolution， engineering applications， key enabling technologies， constraints， and development trends. Recent advances in digital model-driven methods， three-dimensional vision， point cloud recognition， adaptive multi-layer and multi-pass welding， digital twins， and mobile and collaborative robots are also comprehensively analyzed. The review shows that relatively mature dedicated equipment and robotic automation systems have been developed for regular and highly repetitive factory fabrication scenarios， such as steel bridge U-rib panel units and diaphragms. Welding of H-shaped， box-shaped， and stiffened members in building steel structure fabrication is evolving from offline programming toward BIM- and CAD-model-driven operation， three-dimensional vision-based recognition， and teaching-free welding. However， complex non-standard joints， thick-plate multi-layer and multi-pass welding， and on-site operations at height or in confined spaces remain major challenges to the large-scale deployment of welding robots. The current technological focus has expanded from simply improving robotic repeatability to online perception of the actual component state， groove variations， thermal deformation， and weld-bead formation， as well as closed-loop decision-making for welding sequence， bead arrangement， process parameters， and weld quality. The technological pathways for robotic welding of steel structures are shifting from relying primarily on component standardization to reduce operational complexity toward enhancing adaptability to non-standard steel structures through improved perception， planning， and adaptive capabilities， while the two pathways are expected to coexist for an extended period. Future research and applications are likely to place greater emphasis on the integration of models， data， and welding knowledge； multimodal quality perception； autonomous mobility， human–robot collaboration， and multi-robot parallel operation； and task understanding， knowledge retrieval， and continual learning enabled by large-scale AI models and embodied intelligence. Nevertheless， welding process constraints， quality verification， data standards， and safety certification will remain fundamental boundaries for the engineering applications of autonomous welding systems.