© SOMA Project

16 November 2018

Prof. Dr. Oliver Brock from TU Berlin and his futuristic vision of the intelligent robot.

A soft part made of silicon, pumped up with an air pump, similar to a balloon - the robotic hand of the future does not look spectacular at first glance. Only the fact that there is a computer programme running in the background that controls the air supply, could perhaps be described as being extraordinary. However, the development having been taken forward by the TU Berlin in cooperation with the University of Pisa, is one of the latest revolutions in 50 years of a robotics history. While robotic hands had been manufactured from metal at high cost and were able to grip hard objects thanks to complex algorithms, the idea of ‘Soft Manipulation’ - in short ‘SOMA’ is extremely new: Robot-systems are to interact with their environment in a way that copies the human handling of everyday objects. The innocent-looking robotic hand made of silicon is so soft and gentle that it can grip objects without causing any damage or leaving dents or scratches. This is exactly why it can be used much more flexible as compared to the so far metallic hands which cost thousands of Euros to manufacture and distinctly more than the silicon alternatives with 300 - 400 Euros. This potential has also been recognised by the Italian Institute of Technology (IIT) in Genoa, the Deutsche Zentrum für Luft- und Raumfahrt (DLR)[German centre for air travel and space journeys] in Munich, the Institute of Science and Technology Austria (IST Austria) as well as both industrial partners Ocado, a British online-supermarket , and Disney Research in Zurich, who are also involved in the TU Berlin project and are testing the soft robotic hand in practice.

Prof. Dr. Oliver Brock, TU Berlin © SCIoI

At the beginning there was a hand...

The robotic hand that was presented in Cannes in 2014 and whose further development has been supported by the European Union with about 80 billion Euros for a period of seven years, is only one step on a long journey. Or more than that - a means to an end.  

“I believe that we will be in a position sometime in the future to build robots and machines that will behave intelligently”, explains Professor  Dr. Oliver Brock, who is responsible for the ‘SOMA’ project at the TU, “ we programme robots, screw hands and sensors to robots, place them in public and watch how they behave. Our aim is quite clear: We intend to improve the robot behaviour until we have reached a point where we could say that the robot behaves intelligently.“ The vision of the computer scientist who was appointed ’Humboldt-Professor‘ in 2009, seems radical: Robots that are nearly as intelligent as human beings and at the same time act incomparably more precisely. While machines so far have been designed rather one-dimensional and are to perform one specific recurring task in the same surrounding, the new generation of learning machines is to manage complicated reactions and movement processes. In this way the autonomous helpers of tomorrow could be made use of for example in medicine, space travel, or in case of radioactive accidents. To this purpose, the spokesman of the excellence cluster ‘Science of Intelligence’ and Head of the Robotics and Biology Laboratory at the TU Berlin, equips robots with cameras and delicate gripping arms and develops computer programmes for them. Self-learning algorithms give machines a brain in a way. They learn to make independent decisions.

Intelligence versus Artificial Intelligence

“This is a long and time-consuming task”, admits Professor Dr. Brock, who, before returning to Germany in 2009, had done research work in the U.S. for 15 years, and does not entertain any illusions. The reason for this is not only to be found in the robotics itself. In order to understand what accounts for intelligence in human beings and to transfer this to a robot, it is necessary to understand it. “The scientific research locates intelligence in the brain“, says the 49-year old professor, “we recognise intelligence when we see it and can measure it with a test. But even psychologists who are dealing with this professionally, cannot define intelligence.“  He is astonished that science is very much involved in the knowledge itself, however, cannot say how this knowledge is gained.  

It is exactly at this point that the research strategy of the excellence cluster ‘Science of Intelligence’ comes into its own. What fundamental laws and principles are the basis for different forms of intelligence – is it artificial, individual or collective intelligence? Scientists from different disciplines - from psychology via robotics, computer sciences right up to philosophy and ethology - use their research results in order to create new intelligent technologies. The methodical strategy of the mutual cluster of the TU Berlin and the Humboldt-Universität zu Berlin within the excellence strategy of Federal Government and the federal countries is a new approach towards the research of intelligence, where all findings, methods, concepts, and theories must be entered into technological artefacts, for example robots or computer programmes. These artefacts serve as a mutual ‘language’ that is to enable a scientific exchange across disciplinary boundaries.

While watching human abilities - observing, moving and thinking - one fact has become obvious to the computer scientist who studied in Berlin and Stanford:  

“We humans do not solve a problem in its entirety but only small parts of it”, he explains, “and even that happens with clever tricks which we have taught ourselves in the course of evolution. We must understand these tricks to be able to teach them to the robots.”

About Seeing and Gripping

The gripping of the human hand is one of these tricks. We grip things and immediately interact with our surrounding in order to get to the desired object fast and without problems. Turning the page of a book, opening a screw cap of a bottle, an index finger that becomes an improvised in shoehorn. All these are movements that we carry out every day with an enormous efficiency and skill without thinking about it. The ‘SOMA‘ hand has taken the human original as an example. Instead of equipping the robotic hand with sensors in a complicated manner, one only used an air pump - the hardware itself was outsourced. “Computer programmes are running in the background that take care that the environment is no longer considered an obstacle”, according to Professor Dr. Brock, “but a means to achieve a goal faster and easier; just as the humans are doing.“

For the father of robots the eye is also fascinating: “Each eye has 130 million photo-receptors that perceive light“,  he says quite enthusiastically, “however, only about 500,000 bits of information are transmitted to the brain. But instead of this reduction we believe to perceive our environment completely. How does that happen without relevant information getting lost? This is done by a trick.“ Together with his team at the Technische Universität Berlin he equips robots with cameras and tries to understand this trick. And successfully: The robots are really able to select information that is required to fulfil their task. “Of course, neither the tasks nor the eyes are so complex compared with us humans”, admits Professor Dr. Brock modestly, “but the information reduction factor of the machines can nearly be compared with that of human beings.” However, the experiment revealed something quite important: The reduction also simplifies fulfilling subsequent tasks. This was observed in case of the artificial eye as well as also in the robotic hand.

“Perhaps intelligence is a summary, an interaction of these tricks that help each other to solve problems. Science of Intelligence intends to solve this and other questions and also to find new connections in order to understand intelligence and to fundamentally bring it forward“, concludes the scientist who received many awards. It is a vision of the future of a new generation of intelligent robots and artificial intelligence, where Professor Dr. Oliver Brock sees a great potential. “Four eyes see more than two”, the expert for machine learning is quite optimistic, “perhaps two intelligences - the human and the machine - think better than just one. Perhaps it is possible that many decisions we have decided wrongly in the course of history, could be improved in future with the assistance of machine intelligence. I look forward calmly to a future where we have intelligent machines at our side.