Search the laboratory

A mobile platform for studying visual perception, task planning and coordinated mobile-manipulator operation in constrained welding spaces.

3D scanning, groove reconstruction, planning and robot experiments support validation of multi-pass welding for saddle-shaped joints.

Weld-pool imaging, arc acoustics and electrical measurements support process-state recognition and welding-quality research.

RWIA 2026 and CCRW 2026 will take place in Shanghai on December 11–13, focusing on AI-enabled robotic welding. Abstracts are due November 30; online registration is scheduled to open November 15.

Academic presentations, technology exchange and a technical visit addressed robotic welding, intelligence and automation. Professor Shanben Chen chaired the conference; over 200 delegates attended.

Shanben Chen founded the laboratory in March 2000 and continues to lead its research in intelligent robotic welding. His work spans visual and multisensor perception, modelling and intelligent control of welding dynamics, locally autonomous welding robots and multi-robot manufacturing systems.

Huabin Chen studies intelligent welding in complex environments, including multimodal perception, quality monitoring, point-cloud-based path planning and intelligent thermal manufacturing. His work combines process knowledge, data modelling and system integration for welding and additive manufacturing of large-scale equipment.

Yanling Xu researches robot vision, intelligent welding and multisensor quality monitoring. Her work covers teaching-free localization from 3D point clouds, autonomous seam tracking, digital-twin-assisted automatic programming, and robotic wire-arc additive manufacturing and remanufacturing.

Jijin Xu is an Associate Professor and doctoral supervisor at Shanghai Jiao Tong University. His research combines experimental characterization and computational modelling to study welding mechanics, superalloy additive manufacturing, joint strengthening and toughening, and residual-stress measurement and control.

Chun Yu holds a PhD in Engineering and is an Associate Researcher and doctoral supervisor at Shanghai Jiao Tong University. His work combines process monitoring and control, microstructural characterization and numerical simulation for metal additive manufacturing, microelectronic interconnections and welded-interface reliability.

Xiaoqi Chen studies materials processing, intelligent manufacturing and specialized robotics. He was appointed as an SJTU visiting professor in 2011 and served as a chair professor in Materials Science and Engineering from 2015 to 2021, collaborating with the laboratory in research and education.

Na Lv is an associate research professor at SJTU's School of Automation and Intelligent Sensing. Her research covers multisensor information, signal processing and online quality monitoring. She completed doctoral research on intelligent robotic welding under Shanben Chen, with academic links to welding acoustic sensing and multisensor monitoring.

Yiyang Zhao is a postdoctoral researcher at the Institute of Welding and Laser Manufacturing, School of Materials Science and Engineering, Shanghai Jiao Tong University, where he obtained his doctorate. His research focuses on wall-climbing welding robots and intelligent welding-process monitoring, including robot motion control, weld formation quality and intelligent welding equipment.

Yingcai Wan holds a doctorate in engineering and conducted postdoctoral research at the Institute of Robotics, Shanghai Jiao Tong University, from July 2024 to September 2026. His work focuses on robot perception, stereo vision and 3D reconstruction, embodied intelligence, and autonomous mobile welding for large steel structures.

Kai Liu studies 3D visual perception, point-cloud processing and adaptive robotic welding path planning. His work includes vision algorithms, robot hardware–software integration, process experiments and data modelling for intelligent welding in energy and advanced equipment manufacturing.

Bin Zhong studies intelligent sensing and quality diagnosis in welding, combining process mechanisms, domain knowledge and multimodal data for defect recognition. His work includes multisource data acquisition, diagnostic models and online monitoring, with a focus on defect warning and quality traceability.

Jiehao Shen studies laser directed energy deposition of medium-entropy alloys. He participates in the national R&D project on damage suppression and shape–property control in additive, strengthening and subtractive hybrid manufacturing (2022YFB4602102), working on multi-axis path planning, process experiments, multisensor monitoring and process databases for aerospace components such as liquid-rocket-engine turbopumps.

Ziyu Tian studies laser additive manufacturing of nickel-based superalloys and nickel/copper dissimilar materials, focusing on crack suppression and interface control. His work examines process optimization, microstructural evolution and component properties for aerospace and energy applications.

Ming Zhu studies visual perception, weld geometry modelling and tracking, multilayer multipass path planning and process decision-making for complex structures such as large pressure steel pipes. His research connects perception, planning and execution through system integration and experiments, while exploring embodied intelligence and multimodal perception for autonomous welding.

Yunpeng Wang studies mobile manipulator coordination, focusing on 3D visual perception, adaptive robot trajectory planning and coordinated control. His work includes prototype development, process experiments, performance characterization and data analysis.

Yu Han studies welding-process monitoring through multisensor information fusion, with work on algorithm development and process experiments.

Jiahao Gao studies multisensor welding-process monitoring, covering pre-weld 3D reconstruction, in-process weld-pool recognition and post-weld defect detection through algorithm development and process experiments.

Xinyi Li studies multisensor robotic welding.

Zidong Yang studies intelligent prediction of robotic welding process parameters and data-driven modelling. The work focuses on process-data augmentation, ensemble prediction and prediction software development under limited-data conditions, alongside participation in the group’s research and industry collaboration projects.

Liangqiu Zhong studies welding-image information extraction, applying machine vision and deep learning to visual feature extraction, image analysis and welding inspection.

Research on online defect monitoring and coordinated process optimization through multimodal welding-process sensing.

A magnetic mobile platform, manipulator and vision sensors support research on seam perception, path planning and welding control inside pressure pipes.

Development of groove and weld modeling, a welding-process database and dynamic multilayer, multipass planning for pumped-storage pressure pipes, with staged validation using magnetic crawling equipment.

3D scanning, groove reconstruction and path planning support robotic multipass welding experiments on complex intersecting joints.

A magnetic rail, laser seam tracking and weld-pool monitoring support multipass planning, closed-loop process control and on-site welding trials.

How can heat-transfer physics predict penetration depth and weld width when heat input changes over time and the plate has finite boundaries?

How can stereo images with reflections, weak texture and similar edges be converted into usable pre-weld robot paths?

How can multiple partial scans of an intersecting-pipe joint be combined into a continuous welding path with suitable torch orientations?

How can multiple pass trajectories and locally appropriate welding speeds be generated together for a saddle-shaped groove?

How can multimodal process signals reveal defects during robotic welding of excavator booms?

How can process observations support fast temperature-field estimates when full-field measurement is difficult and finite-element simulation is slow?

How can a saddle-shaped header-to-tube groove with varying cross-sections be filled in multiple passes without teaching every point?

How can the back-side pool be observed inside a pipe to recognize penetration during all-position TIG welding?

How can sensing, knowledge-based decisions and process control be integrated for complex, changing welding conditions?

How can heat entering a ground workpiece be tracked as belt wear and contact force change?

How can the surface left by belt grinding be linked to subsequent hardening burnishing to predict final roughness in Inconel 718?

How is grinding energy distributed among the workpiece, belt, chips and surroundings, and how does heat accumulate in abrasive grains?

How can welding states such as penetration, burn-through and misalignment be recognized when new labeled examples are scarce?

Why does material removal change with belt wear even under constant normal force, and can single-grain mechanics explain it?

Can interlayer material design reduce cracking during directed energy deposition of crack-sensitive Inconel 738?

Can grinding sound estimate the remaining cutting capability of a worn belt without stopping it for inspection?

Can ordered, smooth tool-contact paths be generated from scanned point-cloud regions without first reconstructing a complete CAD surface?

How should image-analysis methods be matched to low contrast, interference and subtle defects in welding radiography?

How do thermal softening and elastic belt contact change the balance of rubbing, ploughing and cutting and the resulting heat flow?

Is the current frontside weld-pool image determined only by instantaneous penetration, or does it also depend on the preceding welding history?

How can visually similar laser-welding states be distinguished from subtle differences in molten-pool and keyhole images?

How can welding-state monitoring retain the benefits of an ensemble without running several large networks online?

How can changing pool contours be segmented and categorized as welding pass and torch posture vary in multi-layer MAG welding?

Can ultrasonic assistance and heat-input adjustment jointly improve microstructure and mechanical properties while suppressing cracking in laser-deposited Inconel 738?

How do melt-pool geometry, temperature and sidewall strain affect the stability of laser-deposited thin walls?

Can polishing improve a belt-ground surface without removing most of its strengthened layer?

How can measured geometry guide multilayer, multipass welding of a saddle-shaped header-to-nozzle joint?

How does melt-pool overflow change layer height during multilayer laser deposition, and how can it support prediction?

Can offline planning reduce corner bulging in laser deposition while predicting the resulting layer height?

How can graded stainless-steel/nickel-alloy deposits achieve a better combination of microstructure, strength and ductility?

Can changing pool images, sound and electrical signals predict an abnormal welding state before it develops?

Which optical signatures can be observed during laser welding, and how can they be connected to defects and weld quality?

Can arc sound and electrical parameters jointly identify insufficient shielding gas and weld leakage in CMT lap welding?

Can layer-by-layer laser-power adjustment mitigate heat accumulation and improve the Laves-phase morphology of deposited Inconel 718 thin walls?

How can worn honeycomb structures be removed from distorted turbine seals without excessive damage to the reusable substrate?

How can measured point clouds directly produce practical grinding paths for weld beads on curved surfaces?

Which sensing and control capabilities are needed to move robotic welding beyond teaching and playback?

Sensing, modeling and closed-loop control of welding processes.

Flexible robotic welding cells and coordinated multi-robot operation.

Seam and weld-pool sensing, multimodal acquisition and 3D reconstruction.

Integration of welding paths, processes and equipment for complex components.

Virtual modeling and simulation of robotic workspaces, motion and manufacturing systems.

Autonomous sensing, planning, execution and development of intelligent welding robots.

Remote operation, state feedback and human–robot coordination for welding.

Knowledge representation, machine learning and intelligent decision-making for welding.

Knowledge modeling, data–physics integration and intelligent control of complex welding processes.