The four-legged Mini Cheetah robot from MIT just got one step closer to earning its epic name, thanks to a style of machine learning that’s reminiscent of Neo in The Matrix.
With an unnatural looking gait that only a machine could devise, the robot is now capable of running 8 to 9 miles per hour, a significant increase from its top speed of about 5.5 mph back in 2019. So what changed?
Go, cat, go. MIT’s Mini Cheetah quadruped robot.
MIT
Like Neo and Morpheus practicing kung fu in a simulated training environment, the Mini Cheetah was able to take what it learned from simulated trial and error and apply that knowledge to the real world, making it faster and more capable of traversing different terrains.
“We can train the robot to learn from its experience very quickly in simulation, and then deploy it in the real world,” said MIT Assistant Professor Pulkit Agrawal.
Fetch. The Mini Cheetah mid-sprint.
MIT
Next, the researchers are planning to use similar machine learning techniques to teach the Mini Cheetah how to manipulate objects in its environment.
To see the Mini Cheetah in action, check out the video in this article.
Watch this:
Teaching MIT’s Mini Cheetah How to Sprint
3:24
