13 December 2018

In a recently concluded project in Horizon 2020, researchers and developers from three European countries have created new technology for a disaster-response robot. The robot is intended for use in dangerous environments, such as after an accident in a nuclear power plant.

The Centauro robot
The Centauro robot Photographer: Centauro project
The robot progresses slowly forwards on its four wheeled legs. It’s certainly not in a hurry, but it manages to climb stairs and scramble over obstacles, albeit with a certain degree of wobble. The precision of the robot hands is more impressive: the Centauro robot can grip a freely chosen tool and use it completely autonomously, and it can turn a tap, press a door handle down, and open a door without problems. The robot otherwise is controlled remotely with the aid of an exoskeleton where the operator sits inside the control system and transfers motion to the robot hands, feet and body.

Avdelningen för datorseende, ISY, professor Michael FelsbergMichael Felsberg Photo credit Göran Billeson“Our German colleague who has coordinated the project claims that this is the most advanced system of this type in the world. I’m sure he’s right – not many people have worked on this type of robot”, says Michael Felsberg, professor of computer vision, and, together with Klas Nordberg, head of the LiU contribution to the Centauro project.

Challenging environment

The springboard for the project was the accident in Fukushima, where emergency personnel desperately tried to shut down the damaged reactors with the aid of robots. It didn’t work particularly well. One of the research projects funded by the EU Horizon 2020 initiative, therefore, was to solve the technical challenges and find a way to control a robot that could perform several tasks in a challenging environment where walls and ceilings may have collapsed. The project started in April 2015.

Researchers from three European countries: Germany, Italy and Sweden, have been involved in the project. It was coordinated by the University of Bonn, and the German nuclear safety company Kerntechnische Hilfsdienst GmbH also participated, in the role of end user.
The Swedish participants were researchers from the Royal Institute of Technology and Computer Vision Laboratory, CVL, at LiU.

“At LiU, we have principally worked with perception for navigation and manipulation”, says Michael Felsberg.

Mapping in different scales

The robot can, for example, construct a map of the surroundings, not only at an overall scale, but also at several different scales down to very small features such as small obstacles on the floor. With respect to manipulation, the project has examined how the hand is to be operated such that it can grip, twist, push and pull. One part of the project has also concerned classifying what the robot sees: walls, ceilings, electrical cabinets, furniture and many other features.

“We have also developed a new algorithm for sensors, such as the Microsoft Kinect, that are sensitive to depth”, says Michael Felsberg.

There is much that he is satisfied with, but not everything.

“We’ve achieved a lot, but we haven’t managed to solve it completely. For example, the robot can’t climb up steep ramps: the friction at the front legs is too low for this. It can climb across a ditch, but it can’t do it quickly, which is something that a dog manages with never-failing precision”, he says.

Stronger than us

The robot is not adapted for radioactive environments either, but otherwise works well. The Centauro robot weighs only 75 kilos, but its arms are significantly stronger than a Centauro en robot utvecklad för farliga miljöerThe Centauro robot can grip a freely chosen tool and use it person’s arms, and it has greater stamina. It can hold ten kilos in each hand with outstretched arms indefinitely, and it can keep its hand at exactly the same point above its head while the rest of its body moves.

“But as in all EU projects, traveling takes up a lot of time, and a lot of reports have to be submitted. The projects are also strictly controlled, which makes it difficult to reach true scientific breakthroughs”, says Michael Felsberg.

The Centauro project has, however, despite this resulted in around ten scientific articles for the researchers at the Computer Vision Laboratory.

Translation George Farrants

More information about Centauro (external link)

Video

Contact

CVL news

Tomas Landelius and Carolina Natel de Moura.

The focus period resulted in new collaborations for the climate

In the fall of 2024, researchers from around the world once again gathered at Linköping University for ELLIIT's five-week focus period. This time, the goal was to initiate and deepen collaborations in climate research using machine learning.

Participants are listening to a lecture.

Symposium aiming to improve the climate

In the fall of 2024, Linköping University once again hosted ELLIIT's five-week-long focus period. This guest researcher program aimed for greater breadth in interdisciplinarity this year, with the theme of machine learning for climate science.

Two men and a woman talk in front of a screen

Machine learning can give the climate a chance

Machine learning can help us discover new patterns and better tackle the climate crisis. Researchers from all over the world meet at Linköping University with the goal of finding and deepening collaborations in this area.

Latest news from LiU

Server room and data on black background.

Machine Psychology – a bridge to general AI

AI that is as intelligent as humans may become possible thanks to psychological learning models, combined with certain types of AI. This is the conclusion of Robert Johansson, who in his dissertation has developed the concept of Machine Psychology.

Research for a sustainable future awarded almost SEK 20 million grant

An unexpected collaboration between materials science and behavioural science. The development of better and more useful services to tackle climate change. Two projects at LiU are to receive support from the Marianne and Marcus Wallenberg Foundation.

Innovative idea for more effective cancer treatments rewarded

Lisa Menacher has been awarded the 2024 Christer Gilén Scholarship in statistics and machine learning for her master’s thesis. She utilised machine learning in an effort to make the selection of cancer treatments more effective.