Cyber Valley – from basic AI research to application

"This article was published in our February 2019 newsletter". Sign up here.

Silicon Valley in Baden-Württemberg in southern Germany? Not quite. But it is no coincidence that the Cyber Valley research network was named after the tech region on the US West Coast. The Swabian centre is located in the valley along the Neckar river between Stuttgart and Tübingen. The two cities, together with their universities and the Max Planck Institute for Intelligent Systems, form a hot spot for research in the field of artificial intelligence (AI). Moreover, this is also an outstanding region for industry and home to technology giants such as Daimler and Bosch.

Giving robots the ability to learn

Cyber Valley – from basic AI research to application
As yet, AI researchers can only dream of self-adapting robots.

One of the research groups in Cyber Valley is headed by Jörg Stückler at the Max Planck Institute for Intelligent Systems in Tübingen. A computer scientist, Stückler is an expert when it comes to robotic perception and interaction capabilities. Together with his team in the Embodied Vision research group, his goal is to develop software systems that will allow robots to be programmed such that they no longer have to be designed for specific tasks. "In conventional robotic systems, software is subdivided into different components, each being developed for a specific application", explains Stückler. For example, one component – that is to say a software program – might process sensor data to enable the robot to recognise objects, while a different component would allow it to interpret movements.

From machine-based perception to understanding

"We are exploring a holistic approach. To this end, we are combining perception components and movement components in a learning system", continues Stückler. His research group believes that software structures such as neural networks and deep learning are the way forward; by allowing autonomous learning, deep learning already makes things such as language recognition and Google's AlphaGo games robot possible.

Further developing software structures

The challenge facing Stückler is to further develop these technologies to the point where robots can use data about their surroundings made available to them by video cameras or pressure-sensitive tactile sensors not only to classify objects but also to take decisions. "With the sensor data the robots shall observe the environment in such a way that they are able to predict the effects that actions will have on objects. In turn, they can apply this model knowledge they have learnt to control their own movements", explains Stückler. The aim is to design a robot capable of adapting to new environments, objects and changing situations. "Our hope is that this will not only make robots flexible in terms of how they are used, but will also allow them to communicate and learn from one another."

Ten AI research groups established in two years

Cyber Valley has suddenly turned this region in the southwest of Germany into a hub for one of Europe's largest AI research networks. Funded by the Max Planck Society, the Universities of Stuttgart and Tübingen, and by a number of foundations and companies, ten research groups in the areas of machine learning, computer vision and robotics have already begun their work in the two years since Cyber Valley was founded. Two new chairs in machine learning at the University of Tübingen (only in German) have also been established.

A finger on the pulse of theory and practice

According to Jörg Stückler, the Max Planck Institute in Tübingen and the Cyber Valley research network offer outstanding conditions for his research. "I have excellent colleagues here, as well as many opportunities to collaborate with researchers at the universities in Tübingen, Stuttgart and Zurich. I also have the chance to find out from companies what the requirements are for various applications, and where they see a need for research."

Promoting young researchers at Cyber Valley

The computer scientist and robotics expert belongs to the first generation of Cyber Valley researchers. Many more could follow thanks to the research network's graduate school, the International Max Planck Research School (IMPRS) for Intelligent Systems, where the first 100 PhD students are currently being trained. And who knows, perhaps Jörg Stückler will decide to start up his own business one day. After all, machine intelligence for universal use – known as Artificial General Intelligence (AGI) – is the great dream of AI researchers.

VIDETE

Teaching machines how to see and understand is also the goal of the VIDETE research project pursued by the Augmented Vision department at the German Research Center for Artificial Intelligence (DFKI) in Kaiserslautern. The aim is to programme vision-specialised computers such that they can use images supplied to them by a single camera to reliably recognise moving objects and assess situations. Ultimately, the necessary calculations need to be performed so efficiently that they can also be done by devices with small processors, such as an endoscopy camera for example.

www.dfki.de > VIDETE