Recent Progress Toward Intelligent Robot Assistants for Non-Destructive Disassembly

One key to successful and fluent human-robot collaboration in disassembly processes is equipping the robot systems with greater autonomy and intelligence. In this paper, we present our progress in developing such an intelligent robot assistant system. We present the multi-agent control architecture we developed and describe its technical implementation. Our control approach relies on two types of knowledge models: product and process models. The product model describes the structure of the product to be dismantled through a description of the parts involved and the connections between them. A connection type-related process model describes which agents and objects participate in the disassembly process and which methods of human-robot cooperation or collaboration are most useful. When supplied with a product model and a goal definition, our robotic assistant can automatically generate a partially ordered disassembly task sequence. For each disassembly task, the user can choose one of the divisions of labor defined in the process model. Using the process model, through perception and deliberation, the control system can both execute actions and coordinate and synchronize the actions of human and machine. This approach allows the system to be more autonomous when providing assistance to human coworkers in complex and one-piece disassembly processes.