Together, the robots of the future should be strong. They should discuss and distribute tasks in such a way that they reach the goals that people gave them as quickly and as well as possible. But they must learn to understand that.
Researchers at the Karlsruhe Institute of Technology (KIT ) have selected-first of all odd-looking role models for their work on a new generation of robots: The Seven Samurai, the former football players Manfred Kaltz and Horst Hrubesch, the Rolling Stones and Marvel’s Avengers. But what irritates at first glance, can be explained. All role models were strong only with others. They were able to overcome challenges and achieve goals in which the individual, despite his outstanding abilities, would have failed. The same should be done by the robot teams that the researchers have in mind. In the deep sea or on extraterrestrial expeditions, they should bundle their skills and agree to perform their tasks largely autonomously.
Robots will work independently on Mars
“Autonomous robot networks will become a key technology in the future “, says Michael Flad, Research Group Leader Cooperative Systems at the Institute for Control Systems (IRS) of KIT. For example, they should be able to independently build up systems and take samples on Mars or the ocean floor. Floating or flying drones should scout appropriate sites for equipment or surveys while other units move there to unload or transport material.
“For such tasks, we need robotic systems with multiple capabilities,” says Flad. For example, they need to be able to grab and manipulate unknown objects, to adapt to changing environments, and, above all, to distribute the tasks among themselves. In doing so, one has to take into account the different capabilities of the robots.
US robots can read and implement illustrated instructions
The developers from Karlsruhe could team up with those from startup Vicarious AI from San Francisco. They have developed robots with two arms that can be equipped with different grippers. In addition, they are able to interpret series of pictures that represent the assembly of furniture, Lego bricks or technical equipment, and translate into movements. So far, robots are at a loss for such images.
“Even if you try to teach a computer by pretending something, it repeats the movements exactly,” says Dileep George, Artificial Intelligence specialist at Vicarious. “He does not understand the concept behind it.” Georges latest development already. The robot can recognize concepts and, just like humans learn, 500 has already been saved. For large tasks such as the construction of furniture or Lego figures, however, it is not enough. After all, he can already arrange figures for a pictorial template, which are confused. “Our work brings us closer to the goal of building robots that have some sort of common sense,” says George.
Nuclear Decommissioning and Autonomous Driving Networks
That Karlsruhe network is not only suitable for pure research when it comes into action. It could also be used in autonomous driving as well as in logistics and transport. Even in disaster areas, the smart robots could be used. Or in the dismantling of nuclear plants. So far, there are only remote-controlled mechanical helpers in both cases. However, if operators can not assess the situation, perhaps because smoke makes the transmission of images impossible, these robots are no longer useful.
There is great potential for machines that negotiate and act together. “Without cooperation between the autonomous systems and interactions with humans, which will be present at least in the medium term as an actor, however, automatic driving can bring no plus to road safety, time savings or comfort,” says Flad.