DLR Logo
 
Fontsize: [-] Text [+]
Department of Robotic Systems

ICRA2011

Modular state-based behavior control for safe human-robot interaction: A lightweight control architecture for a lightweight robot




New video

Playing basketball with the DLR Lightweight Robot III



- SAFE-ROBOTS -

 


Team     -    

Publications

    -     Videos




At the Institute of Robotics and Mechatronics of the German Aerospace Center (DLR) a new generation of safe robots was developed for direct physical cooperation with humans.This will qualify the robots as assistants for humans in industrial assembly and manufacturing, medicine, or, one day, even as a house-hold help in everyone’s home. In all these applications it has to be ensured that the robot will under no circumstances injure the humans due to malfunction or even in case of user errors. This research tries to systematically study and classify all possible injuries which can appear in human-robot interaction, generate a large experimental data basis for understanding and evaluating these potential injuries and provide design and software measured to avoid or at least diminish them as much as possible.

The DLR robots are characterized by their lightweight structure and programmable compliance which is based on measuring the interaction forces and torques (Soft Robotics) using force-torque sensors. This feature allows executing interaction tasks with humans and the environment in a skillful way. In order to accomplish tasks in everyday environments, the robot strongly relies on its compliant behavior to accommodate for position errors on. 

Furthermore, the force-torque measurement enables the robot to detect collisions and react effectively in order to avoid injuries. This method was tested exhaustively at standardized crash test facilities at the German Automobile Club as well as with mechanical devices at DLR. The outcome of these tests indicated that this new generation of robots is intrinsically not able to produce any significant injury by means of well established injury indicators used in automobile crash testing. In other words the obtained impacts fall far below any safety-critical threshold of the evaluated severity indices. In order to ultimatively prove this strong statement we conducted a one-time series of human-robot impacts (please note that the subject was the first author of the studies) including the head, chest, shoulder and abdomen. During these experiments we proved the feasibility, reliability and effectiveness of our collision detection and reaction strategies. While fulfilling its task the robot is able to distinguish between desired physical cooperation and accidental collision. The latter triggers the robot e.g. to stop the motion and react compliantly when pushed away by the human, therefore posing no threat. 

Based on these results we evaluated (not by human impacts!) several industrial robots ranging up to a weight of 2500kg. We showed that generally blunt head or chest impacts without the possibility of a human being clamped are, no matter how massive the robot is, definitely not life-threatening. On the other hand in case of clamping the situation is fundamentally different. While the new generation of DLR-robots is still not able to exceed the above mentioned safety-critical thresholds, industrial robots are potentially lethal. 

Apart from blunt impacts it is of even higher importance to make robots safe also for sharp contact. Future systems are supposed to work e.g. as industrial co-workers or service robots. In order to fulfill such demanding tasks the robot will make use of a variety of tools as humans do in countless everyday situations. Humans are able to handle extremely dangerous tools without posing a threat to other humans. To equip robots with exactly the same capabilities of safe interaction even for such highly delicate tasks (and making robots even safer than humans) is the aim of a series of soft-tissue impacts we carried out. We were able to extract the relevant information about the mechanism of such injury and design according countermeasures by means of collision detection and appropriate reaction. We were able to show that due to its force sensing capabilities the robot is able to reduce or even entirely prevent injury

The described work is part of various projects as e.g. the European project PHRIENDS (Physical Human-Robot Interaction – Dependability and Safety) and VIACTORS (Variable Impedance ACTuation systems embodying advanced interaction behaviORS). 

For further information please contact:

Sami Haddadin

Alin-Albu-Schäffer