Operating a Robot in Germany from the USA?

Operating a Robot in Germany from the USA?


CITEC researchers are developing a system together with Indiana University (USA) that enables robotic experiments to be replicated anywhere, regardless of location

The reliable and precise reproducibility of scientific experiments is crucial in research. In the natural sciences, such as biology, chemistry, and physics, it is an established practice to design an experiment so that it can be reproduced by other scientists. In robotics, however, reproducibility is still a challenging topic. Computer science doctoral researcher Florian Lier is developing in cooperation with the Central Lab Facilities and the Cognitive System Engineering research group an approach that facilitates the replication of software-intensive robotics experiments. The developed concept will be evaluated and improved in the Thematic Network Interactive Intelligent Systems together with researchers at Indiana University.

Robot experiments can be reproduced worldwide with the Cognitive Interaction Toolkit. Photo: CITEC/Bielefeld University “In an experiment, when a robot takes an object in its hand and is supposed to pass it on to a person, various technical and methodical criteria must be considered if an experiment is to be reproduced at a different location,” says Florien Lier, who investigates the topic of the reproducibility of software-intensive robot experiments in his dissertation.

“Research findings in robotics are typically published as classic written articles in which hypotheses are proven or disproven using purely formal mathematical or data-driven statistical analyses,” says Lier. “Basic questions we ask ourselves not only in robotics, but in all computer-aided sciences include: how does the software used affect the analysis of the data? How much do the published findings depend on the software and hardware configuration used and the environment of the experiments? Can the same results also be attained using a form of the software that has only been slightly modified? In order to find answers to these questions, Lier is developing with his CITEC colleagues the Cognitive Interaction Toolkit (CITK).

The aim of the CITK is to achieve full reproducibility of a robotic experiment. In addition to the technical reproduction of the experimental system, the CITK provides a development process, which supports the transparent reproducibility of partial findings all the way through the final data set. This makes it easier to digitally share the basis for research findings, and to build further developments upon existing knowledge. Part of the toolkit includes a catalogue and various archives in which projects that have already been run are stored, and are available to any researcher around the world, just as in a library: http://toolkit.cit-ec.uni-bielefeld.de/.

“We have created step-by-step instructions that allow researchers to quickly and easily download the data and create an identical software environment – that is, a part of the experiment’s set up,” explains Florian Lier. Accordingly, a research group in the United States can, for instance, create the technical foundation for a robot experiment in the same way a group in Bielefeld can.

In addition to the technical aspects, there are also of course human aspects: The platform “jspsych” regulates what the study participants should do in an experiment. Phillip Lücking, who works in CITEC’s Central Labs, transferred this platform from the department of psychology at Indiana University to experimental set-ups in robotic research. The platform “jspsych” regulates the timing of the test procedure, or the display of stimuli, or for instance activates a certain type of robot behavior.

Together with Robert Goldstone and Selma Šabanović from Indiana University, the CITEC researchers currently planning to test their approach of reproducible researching in an upcoming experiment from psychology and human-machine. To do this, the researchers will make an experiment reproducible with the help of their toolkit. This will allow them to test whether the platform can be used publicly by international researchers. On 15 May 2017, Bielefeld researchers will travel to Indiana University to carry out a set of experiments with their colleagues in the USA.

In addition to the scientific value of reproducibility, the toolkit is also useful in this era of globalization and digital internationalization. Researchers will no longer have to travel to different countries to inspect experiments with robots on site. The toolkit already provides a so-called “online replay” for some selected experiments. With this, recorded data and the stages of the experiment can be played back with the click of a button. Researchers can then conveniently follow along the findings in their own Internet browser.

Currently, CITEC researchers are working on expanding the system to make it possible to locally start the systems and experiments gathered in the CITK toolkit. The experiments can then also be used by cooperation partners from other countries in a “remote lab”. In this way, the experiments can be replicated in the United States, Japan, Germany, or anywhere in the world. And what’s more: it is possible to observe an experimental set-up in Germany, or even perform that experiment, from the United States.

Contact

Florian Lier, Central Lab Facilities
Dr.-Ing. Sebastian Wrede, Cognitive Systems Engineering
PD Dr.-Ing. Sven Wachsmuth, Central Lab Facilities
Cluster of Excellence Cognitive Interaction Technology (CITEC)
Bielefeld University
Telephone: 0521 106 293
Email: flier@cit-ec.uni-bielefeld.de

More information is available online at: