The four-legged robot Dyret adjusts the length of the legs to adapt the body to the surface. Along the way, he learns how to get around best.
The name Dyret (Norwegian for “animal”) is an acronym for Dynamic Robot for Embodied Testing.
“We have demonstrated the benefits that a robot can continually adapt its body shape. Our robot proves that this can be done easily with modern technology, ”explains Tonnes Nygaard, Senior Lecturer in the Department of Informatics at UiO.
In Dairet’s case, the reshaping of the body means that he regulates the length of the legs. The body shape adaptation mechanism has proven to be very useful for the robot.
Scientists have previously shown that their robot adapts to different environments under controlled indoor conditions. Nygaard then spent six months with fellow engineers at the Commonwealth Scientific and Industrial Research Organization (CSIRO) in Australia. She specializes in outdoor testing of self-learning robots.
“It used to be considered too difficult to achieve in the real world. With the help of robots and our experiments, we have shown that this is possible, ”explains Nygaard in an interview with Titan.uio.no. The results of the work were published in the journal Nature Machine Intelligence.
By changing the length of its legs, the robot can automatically change the shape of its body. Morphologically adaptive robots can operate in unpredictable environments and face new challenges without having to redesign their structure or rebuild each time they encounter something unexpected.
It is difficult for humans to imagine how difficult it is for a robot to move, for example, from concrete to grass. Do not forget that a person has many years of experience and quite a lot of feelings compared to a robot.
“The robot uses a camera to see how uneven the terrain is, and sensors in its legs to determine how hard the surface is to walk,” explains Nygaard. “Diret is constantly learning about the environment he walks through and, combined with the knowledge gained from indoors in a controlled environment, uses this to adapt his body.”
When Diret was asked to walk on the grass, he had never seen grass before. He only trained on gravel, sand and concrete. However, he quickly learned to walk on Australian grass and what the ideal leg length was. It was found that shorter legs provide better stability for the robot, while longer legs provide higher walking speed if the ground is predictable enough.
A flat lawn may not be the biggest problem, but the grass in nature is full of tufts and holes that can trap a leggy robot, so Diret shortens his legs. On concrete, he can pull them out and “run away”.
The robot also adapts to damage when hitting unforeseen obstacles.
“Using our technology, the robot can adapt to the fact that one of its legs becomes weaker or breaks. He can learn to recover by limping or shortening the other three legs, ”says Nygaard. Learning from her mistakes and breakdowns, Dairet adapts better to her environment.
While Dairet is not ready to take on serious tasks. The goal of Nygaard’s doctorate was to develop technology and find suitable materials, and prove that it is possible. However, he sees several possible future uses. For example, the robot is useful in search and rescue operations, as well as in agriculture, where there is a wide range of difficult surfaces and weather conditions. Also, Dairet will help in the exploration of mines where people are difficult to get.