FlexIRob - Motion Learning at your Fingertips

Enabling robots to become co-workers that collaborate with humans efficiently and safely is a major goal of current robotics research. At Bielefeld University's CoR-Lab researchers use learning and interaction technology for control of compliant robots such as the KUKA Lightweight Robot (LWR) to realize such human-robot interaction. Research results are continuously integrated in the showcase robotic system “FlexIRob” providing a testbed for flexible robotic co-worker and advanced human-robot collaboration scenarios.

In its current setup, FlexIRob allows to teach a redundant robot various nullspace constraints in different areas of the workspace. Users with no particular robotics knowledge can perform this task in physical interaction with the compliant robot, for example in order to reconfigure a working cell environment. After a short training phase, the learned adaptive mapping solves the inverse kinematics problem of the robot. It is embedded in the motion controller of the system, hence allowing for execution of arbitrary motions in task space, respecting the learned nullspace constraints. This is a large step towards our vision of a flexible robotic coworker system, because it avoids the complex manual programming that standard methods for this task have previously required.

For the FlexIRob scenario, the Cognitive Systems Engineering group at CoR-Lab developed a system architecture for the control of compliant robots, in particular for the KUKA LWR, that provides a set of coherent tools for high-level simulation, programming of individual and re-use of complete learning or interaction components. Learning in this system is realized by means of neural-network-based learning algorithms such as the Backpropagation-Decorrelation learning method developed at the CoR-Lab in the Cognitive Robotics research group.

The following videos show the teaching of different redundancy resolutions in different parts of the workspace and the robot executing motions adapted to its environmental constraints.

The FlexIRob setting is used as technical basis in several projects including the it's OWL Human-Machine Interaction, CogIMon, and CoHRoS.

Publications:

Contact:

Prof. Jochen Steil
Dr.-Ing. Sebastian Wrede
Dr.-Ing. Christian Emmerich