NRI:FND: Intelligent Co-robots for Complex Welding Manufacturing through Learning and Generalization of Welders Capabilities

NRI:FND: Corobots with Expert-like Intelligences for Complex Welding Manufacturing YuMing Zhang, Peng Wang Abstract: There is a dramatic and growing shortage of highly skilled welders, accentuated by the fact that manufacturing complexity and production levels are rising. Specialized skills, including adaptive judgement and decision makings, specific to sophisticated operations require many years of practice to develop. While fully robotic systems have been able to replace human welders for relatively simple welding tasks, they are unable to successfully accomplish more complex tasks. Also, development of adaptive control systems for robots is lengthy and costly. To address these challenges, the goal of this project is to integrate process sensing, visualization, and deep learning-enable modeling techniques to formulate a systematic solution that allows us to efficiently advance robotic capabilities on acquiring domain-specific knowledge, interactive learning, adaptive decision making, and collaboration, for realization of fully robotic automation on complex welding processes quickly. As one part of the solution, a virtual welding environment will be created to allow human welders virtually interact with the welding scene with improved safety and comfort, while their adaptive operations can be recorded for knowledge extraction. In addition, the advanced modeling technique will generalize the knowledge extracted from human welders operating on different welding scenes, and hence reduce the information, effort, and cycling time needed for developing a robotic control and learning system.