Robots these days are equipped with locomotion systems to overcome all types of terrain, but tend to favor one type of environment in particular. Scientists at Norway's University of Oslo have instead developed a four-legged robot that can adjust its leg length and walking gait on the fly as it encounters different surfaces, an ability they say improves its energy efficiency and performance in unpredictable settings.
Called Dyret ("The Animal" in Norwegian), the robot is described as the first quadruped robot that can autonomously change its morphology in response to different conditions. By using a mix of sensors, cameras and artificial intelligence, the robot is able to recognize when it encounters different terrain underfoot and mechanically adjust the length of its legs, and in turn its body shape, optimizing its gait for that particular surface.
“The robot continuously learns about the environment it's walking on and, combined with the knowledge it gained indoors in the controlled environment, uses this to adapt its body,” says Tønnes Nygaard, who led the research.
The researchers started by training the robot on gravel, sand and concrete, and then tasked it with walking on grass, which it had never encountered before. While this mightn't seem like the most troublesome terrain, the researchers point out that tufts and holes in grass can pose a real challenge for long-legged robots. Dyret, however, learned to walk on the lawn by quickly determining the ideal leg length.
“Shorter legs give better stability, while longer legs allow for a higher walking speed if the ground is sufficiently predictable,” says team member Kyrre Glette.
The idea here is to develop robots that aren't brought to a standstill when they encounter unexpected terrain, and don't need to be redesigned for different scenarios. Such robots could find use in disaster zones, search and rescue scenarios, agriculture, difficult-to-access settings like mines, or even on other planets. These adaptive abilities could also make robots a whole lot more resilient and durable.
“Using our technology, the robot is able to adapt to one of its legs becoming weaker or breaking," says Nygaard. "It can learn how to recover, whether through limping or decreasing the length of the other three legs."
The team has made the technology open source, hoping to inspire others in the field to take parts of the design to build adaptive robots of their own.
The video below from 2018 shows an early version of Dyret in action, while the new research was published in the journal Nature Machine Intelligence.
Source: University of Oslo
