MIT engineers recently unveiled the capacity their four-legged robotic 'Mini Cheetah' can do. According to the researchers, the robot can move on the ground at a speed of nine miles per hour.

The scholars said that the cheetah robot could also traverse complicated surfaces such as gravel and ice.

MIT Mini Cheetah Robot

(Photo: Bryce Vickmark)
MIT’s new mini cheetah robot is springy, light on its feet, and weighs in at just 20 pounds.

The developers considered the Mini Cheetah as 'virtually indestructible' due to its knowledge acquired from passing through any challenging terrains they have included in the tests. The four-legged machine could also learn to analyze its capabilities through trial and error, similar to how humans learn.

With the simulated technology programmed into the MIT running robot, the machine could harness experience that is worth 100 days of practice. However, the total testing time in the real world took only about three hours.

According to a report by DailyMail, robotic animals were first introduced in the industry almost 25 years ago. Since then, researchers and developers have curated many animal-like machines and humanoids that sampled skills, including walking, jumping, and other practical functions that can benefit humans.

 

In recent years, animal robots still demonstrate their modern-day skills. For example, Boston Dynamic's mini robot dog called Spot presented a dance choreography based on the music of the South Korean boy band 'BTS.'

Much for movements, programming an animal robot requires an extensive amount of effort that could make it run across various types of terrain, the MIT developers said.

ALSO READ: Efficient Cleaning Method for Solar Panels Without Water Use Discovered by MIT Experts


Fastest Animal Robot That Can Move in Range of Terrains

MIT Computer Science & Artificial Intelligence Laboratory (CSAIL) expert and lead developer Gabriel Margolis explained that their team constructed a method in which the cheetah robot's behavior can improve its simulated experience.

The approach allows the system of the machine to fully apply its learned behaviors in the real world, Margolis added.

According to the authors, the environment that the Mini Cheetah perceives in the simulator embedded in its system teaches the machine its potential skills that could be deployed in the terrains and surfaces of the real world. When demonstrating the abilities, it is honed from the simulated program; the robot executes the proper skills needed in real-time.

Margolis explained that programming the robot with running skills is much harder than making it walk, as achieving a faster movement skillset requires the hardware to push to its limits.

This is also part of how the robot cheetah robot responds in real-time in a situation where it should move quickly throughout surfaces plotted with grass or ice. Like humans, the machine could certainly adapt to the environment it is into, whether to run on grass or slow down on slippery grounds of ice.

Fellow MIT expert and co-author of the study Ge Yang said that it is challenging for the team to program the robot to act quickly in every situation possible.

According to the developers, the learning system they relayed into the Mini Cheetah robot was successful that it had already broken its record in one of the running platforms it was tested. The paper for this study, titled "Agile Locomotion via Model-free Learning," will be published in 2022.

RELATED ARTICLE: Flippy's Tortilla Chip-Making Successor Chippy Developed to Deep Fry, Season Chips

Check out more news and information on Robotics in Science Times.