"Robots Learn Not to Clean the Carpet with a Mop"

Three questions for Martin Meier, doctoral researcher at CITEC

Computer scientist Martin Meier is a specialist in ambidextrous mechanical grasping. At CITEC, he works in the Neuroinformatics research group, which is headed by Prof. Dr. Helge Ritter. Recently, he started working with the Ease Collaborative Research Center (SFB 1320), which is located at the University of Bremen. At Ease, robots are learning to act autonomously on abstract instructions. The Cluster of Excellence CITEC at Bielefeld University is participating in Ease as an external partner, and Meier is contributing to a subproject on manual intelligence.

Martin Meier, doctoral researcher at Cluster of Excellence CITEC What is the challenge for you in collaborating with the Ease Collaborative Research Center?
At the moment, I am mainly dealing with how we can take research findings from the lab at CITEC and transfer them in a meaningful way into the context of the Ease Collaborative Research Center. The same applies to the transfer of knowledge from Bremen to Bielefeld – we also want to use, for example, the University of Bremen’s software here in the lab at CITEC. In addition to the organizational challenges, there are also, of course, many questions relating to the work itself. At the end of the first funding period in 2021, humanoid robots should be able to autonomously carry out household tasks such setting the table or preparing simple meals. But it is precisely these challenges and the overall open approach that appeal to me in this scientific work.

What exactly are you and the other researchers at the Ease Collaborative Research Center working on?
We at the Ease Collaborative Research Center are dealing with the question how of robots can recognize which manual action, in which context, is the right one. When the robot is instructed to clean a room, it always depends on the setting: a mop cannot be used to clean a carpet, even if the movement is very similar to the one that is used to operate a vacuum cleaner. Other examples include pouring beverages into certain containers – coffee is poured into a cup, not into a glass. The robot therefore has to know how it must move its arm and hand in order to pour something or clean a surface. At the same time, the robot has to be able to identify the context in which the action is taking place. We are investigating how robots can adroitly and efficiently choose the right action from the many options for action, and then carry out the individual steps in a meaningful sequence.   

What do you bring from the Cluster of Excellence CITEC to the Collaborative Research Center in Bremen?  
In the large-scale Famula project at CITEC, we are working for instance on how a robot can autonomously recognize and grasp objects. In order to address this question, we are researching correct grasping position in particular. Our focus is on fine-motor control, including the issue of how the fingers have to be oriented. In Bremen, our colleagues are conducting research on longer sequences of activity. It’s exciting to see how these research fields are brought together in the Collaborative Research Center. In the spirit of open science, the findings are planned to be made available to other labs via an online platform.

Martin Meier (born 1982) studied Computer Science in the Natural Sciences at Bielefeld University. Since 2010, he has been a research associate in the “Alignment in Communication” Collaborative Research Center (SFB 673) at Bielefeld University, as well as the EU project SARAFun (Smart Assembly Robot with Advanced Functionalities). In his doctoral research, Meier is developing a system that recognizes similarities.

The Collaborative Research Center 1320 Ease (Everyday Activity Science and Engineering) started on 1 July 2017 and has been funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) with a total of 10 million Euro for the four years of its first funding period. The Cluster of Excellence Cognitive Interaction Technology (CITEC) is involved as an external partner on the subproject “Episodic memory for everyday manual activities.”

More information is available online at:
Ease Collaborative Research Center: http://ease-crc.org
Large-scale Famula project: www.cit-ec.de/en/deep-familiarization-and-learning
“Self-Learning Robot Hands” press release (31 May 2017): www.cit-ec.de/de/news/die-selbstlernende-roboterhand