Integration of a reconfigurable robotic workcell for assembly operations in automotive industry

This paper deals with the integration of a flexible, reconfigurable work cell performing assembly of parts in the automotive industry. The unique feature of the developed cell is that it can function in two modes: a) entirely autonomously or b) in cooperation with a human, where the operation of the robot dynamically adapts to human actions. We have implemented technologies for online recognition of human intention and for real-time learning of robust assembly policies to achieve the desired outcome. This challenging goals dictate the integration of modern deep learning algorithms, statistical learning, and compliant robot control into a unique ROS-based robot control system.