Auch Roboter wie diese könnten mit Hilfe von Q-Rock konstruiert werden: DFKI-Robotersystem SherpaTT im Feldtest in der marsähnlichen Wüste von Utah. Foto: © DFKI
Author profile picture

BREMEN, December 15, 2018 – Robots have long been an important part of industrial production. They are also fulfilling more and more functions in logistics, especially in the internal logistics of an industrial company. In the past, they could only be found stationary, but now there are more and more mobile and partly autonomous robots that transport individual components or search complex production plants for errors. However, this also poses new challenges to the development of robots. The German Research Institute for Artificial Intelligence (DFKI) is now investigating in a research project how users without expert knowledge can develop robot systems tailored to their requirements in the future.

Artificial intelligence helps to construct robots

Q-Rock is funded by the Federal Ministry of Education and Research with 3.17 million euros. The aim is to develop a solution with which even small and medium-sized companies can develop robots for their own purposes. “Q-Rock is an important step towards so-called ‘integrated AI solutions’. This approach will also enable people who are not AI or robotics experts to develop and deploy systems tailored to their own needs,” says Professor Frank Kirchner, who heads the DFKI Robotics Innovation Center.

An example of modern specialized systems: This robot system from Grenzebach is used at large airports to load baggage. Photo: TheGrenzebachGroup via Wikimedia Commons.

Q-Rock uses artificial intelligence methods such as structural reasoning and machine learning. It also uses data from a previous project, which created a database for the development of robots. In addition to software modules, this database also contains hardware and behaviour models. The individual components are modularized and can, therefore, be combined within certain limits.

In the end, users should be able to access such a database and configure robots from the offered elements according to their specifications. In Q-Rock, the robot itself will be able to understand its capabilities based on its hardware structure.

Robots that understand themselves

Of course, the first thing the researchers have to do is develop special programs that can do this. And they must first describe the capabilities of subcomponents, i.e. a sensor or a joint, before they can derive the capabilities of an overall system. To put it simply, they need digital models of the individual robot components from which the entire machine is created. The components are not only components such as gripper arms, a motor or sensors but also modular software modules for controlling robot behaviour.

Robots are also used in research, here in the National Genome Project in the USA. These robots move sample containers from one workstation to the next. Photo: Maggie Bartlett, National Human Genome Research Institute via Wikimedia Commons.

Q-Rock should also lead to robots that are capable of understanding their skills based on the hardware they are made of. The robot software would first determine the capabilities of individual components on the basis of a general description and then derive the capabilities of the entire system from this. This would teach the robot what to do. Conversely, these abilities can then also be stored as software modules, which are then also contained in the database.

A user can now combine hardware and software components to create a complete robot system. Special prior knowledge is not required. It is sufficient if he enters his requirements into the database. This way, artificial intelligence can help to build highly specialized robot systems with little effort in the future. These would normally be industrial robots, but the principles could also be applied to the construction of space probes or autonomous exploration robots.