This four-legged robot doesn’t look very graceful. However, it’s deep in thought.
Developed by the University of Oslo and revealed to the public in May 2018, this hesitant machine is called Dynamic Robot for Embodied Testing, or “Dyret”. When it starts walking, algorithms choose randomly from eight possible walking configurations with varying gait and leg length. From these, Dyret can mix and match techniques to find the most effective way to walk. By reading which way works best on the terrain around it, the robot creates and discards strategies for traversing the area smoothly. It can also take into account how much power it has left and optimize for the highest speed.
It’s remarkable because this evolution happens in real time. Self-reconfiguring robots like this could adapt to different locations or different jobs without needing a human to program them beyond the initial algorithms. They would be ideal for jobs in which conditions often change.
As you can see in the video, the decisions involve a lot of trial and error. Keep walking, buddy. You’ll figure it out.