Cooperative Human Robot Object Manipulation via Adaptive learning control (CHROMA)
By developing close physical human robot interaction, role of the robot is elevated from *tool* to *co-worker* and hence robots can be an integral part of modern societies and can improve humans quality of living. In CHROMA, we focus to a specific physical Human Robot interaction problem: manipulation of an object that jointly held by a dyad of a robot and human. To perform efficient cooperative action, the robot must have an understanding of the human´s intentions and limitations, as these are communicated nonverbally through the interaction process. We aim to develop a control strategy for human-robot object co-manipulation that can deal with uncertainties in: a) human impedance characteristics, b) kinematic constraints imposed by the human and c) uncertainties associated with human-robot load sharing and object gravity. We model the human as a virtual joint with kinematic characteristics that reflect human intention and dynamic attributes that correspond to human natural impedance while objects are considered to have unknown mass with unknown distribution. Central to achieving the aforementioned goals is the development of theoretically justified adaptive control strategies,combining advanced performance in terms of manipulation and estimation for learning the unknown parameters of the system. Performance of the developed advanced human-robot object manipulation skills will be demonstrated in real-world scenarios e.g. a robot and a human lifting up and leveling an object.
Jonas Sjöberg (contact)
Full Professor at Chalmers, Electrical Engineering, Systems and control, Mechatronics
Swedish Research Council (VR)
Funding Chalmers participation during 2015–2017