Today’s robots are equipped with motion systems to cross all types of terrain, but tend to favor a particular type of environment. Instead, scientists at the Norwegian University of Oslo have developed a quadrupedal robot that can adjust leg length and gait in flight when it touches different surfaces, an ability they give. that improves its energy efficiency and performance under unpredictable conditions.
Known as Dyret (“Animal” in Norwegian), this robot is described as the first four-legged robot that can automatically change its morphology under different conditions. By using a combination of sensors, cameras and artificial intelligence, the robot can recognize when encountering different terrain underfoot and mechanically adjust the leg length, and in turn, the mechanical shape. its body, optimizing gait for that particular surface.
Tønnes Nygaard, head of the study, said: “The robot continuously learns about the environment it is going through and combines with the knowledge acquired indoors in a controlled environment, using this. to adapt to its body “.
The researchers started by training the robot on gravel, sand, and concrete, and then tasked it to walk on grass, something it has never encountered before. While this may not seem to be the most troublesome terrain, researchers point out that tufts and holes in grass can be a real challenge for long-legged robots. However, Dyret learned to walk on the grass by quickly determining his ideal foot length.
“Shorter legs provide better stability, while longer legs allow higher walking speeds if the ground is predictable,” said Kyrre Glette, a team member.
The idea here is to develop robots that don’t get stuck in unexpected terrain and don’t need to be redesigned for different situations. Such robots can be used in disaster areas, search and rescue scenarios, agriculture, inaccessible installations like mines or even on other planets. These adaptations could also make the robot much more resilient and resilient.
“Using our technology, the robot is able to adapt to whether one of its legs becomes weaker or broken,” Nygaard said. It can learn to recover, whether by limping or reducing the length of the other three legs.
The team has made the technology open source, hoping to inspire others in the field to join parts of the design to build their own adaptive robots.
The video below from 2018 shows the first version of Dyret in action, while the new research is published in the journal. Natural computer intelligence.
DyRET: Dynamic robot for embedded testing
Source: University of Oslo