Artificial Intelligence Assisted (Dis-)Assembly—Persistence and Transformation

Assembly involves the creation of products through the spatial arrangement of components and the temporal coordination of actions, all within the constraints of limited human and machine resources. The objects and actions involved are tightly interrelated, imposing mutual constraints driven by technology, product structure, geometry, and tolerances. It is no wonder that the solution to this hard puzzle of production engineering called for the application of artificial intelligence (AI) techniques from the very beginning of the field. The talk gives a concise overview of enduring achievements in AI-assisted assembly, with a focus on key areas such as representation, reasoning and planning, sensing, execution and recovery, robotics, multi-agent assembly—including human–robot collaboration—and learning. It will also be discussed how AI technologies have transformed assembly from an automated process into a more autonomous one, and how they have extended its scope to include disassembly and remanufacturing. This transformation is apparently being accelerated by the proliferation of generative AI and industrial foundation models: one can expect smarter representations derived from CAD and GDT models as well as more adaptive planning methods operating under uncertainties. Manipulation and robotic skills can be learned from demonstrations or simulation and transferred to novel product variants. Agentic AI can provide new recovery methods and hitherto unseen immersive and collaborative environment for teamwork. The talk will, finally, pose the question whether these advances can deliver what the field has sought for decades—a true industrial-level breakthrough in autonomous (dis-)assembly.