Two-legged robot ‘Cassie’ breaks the 100-meter record for a bipedal droid.

Photo of author

By Creative Media News

A two-legged robot capable of pursuing you on the street may sound like something from the most recent episode of Black Mirror.

Cassie, a two-legged robot, has established a new Guinness World Record for the fastest 100 meters by a bipedal robot. This may not be too far from reality.

Cassie clocked a record-setting time of 24.73 seconds by beginning the run in a standing posture and returning to it without falling.

While that is more than 15 seconds slower than Usain Bolt’s 9.58-second world record for the 100-meter sprint, it is faster than the time most humans would take to jog the same distance.

Two-legged robot 'cassie' breaks the 100-meter record for a bipedal droid.
Two-legged robot 'cassie' breaks the 100-meter record for a bipedal droid.

Professor of robotics at Oregon State University Jonathan Hurst, who spearheaded the robot’s development, regarded it as a “huge turning point.”

He stated, “This may be the first bipedal robot to learn to run, but it won’t be the last.”

The 100-meter record builds on the robot’s previous accomplishments, which include running 5 kilometers in approximately 53 minutes in 2021.

Cassie finished the 5K untethered and on a single battery charge on the Oregon State campus.

The robot has ostrich-like knees and functions without cameras or external sensors – effectively as if blind – relying on machine learning to govern its running stride.

The Guinness attempt was directed by graduate student Devin Crowley, who stated, “Over the past several years, while racing a 5K and also ascending and descending stairs, we have been gaining the knowledge necessary to break this record.”

World record
Two-legged robot 'cassie' breaks the 100-meter record for a bipedal droid.

‘Machine learning algorithms have been used for pattern recognition, such as picture recognition, for a long time, but developing control behaviors for robots is a new and different use.’

While completing a 5K helped to demonstrate Cassie’s dependability and endurance, it did not answer the question of how quickly the robot could run.

Consequently, the research team shifted their emphasis to speed.

Cassie’s training in a simulated environment was shortened to a week using a computer approach known as ‘parallelization.’

This involves several processes and calculations occurring simultaneously, enabling Cassie to undergo a variety of training situations concurrently.

“Cassie is capable of a variety of gaits, but as we specialized her for speed, we began to wonder which gaits were most effective at each speed,” Crowley stated.

This resulted in Cassie’s first optimized running stride and behavior astonishingly comparable to human biomechanics.

Cassie needed to reliably start from a standing posture, run, and then return to a standing position without falling.

“Starting and stopping in a standing position is more challenging than running,” said artificial intelligence professor Alan Fern, who helped create the robot’s ‘brain’ “Starting and stopping in a standing position is comparable to takeoff and landing on an airplane,” he added.

The researchers addressed this issue by switching between two neural networks: one that can run and another that can stand.

Everything was then a matter of encoding the correct timing for these transitions to make them work properly.

This 100-meter performance was the product of a close partnership between mechanical hardware design and advanced artificial intelligence for the control of that hardware.

The development of Cassie was supported by a $1 million grant from the US Defense Advanced Research Projects Agency (DARPA) and the National Science Foundation.

Its 100-meter record, set on 11 May 2022, has been officially recognized as a Guinness World Record.

According to Guinness World Records, “given the realistic nature of the job, establishing this record is a concrete milestone in robot locomotion and real-world competence.”

Hurst added, “I believe such control systems will play a significant role in the future of robotics.”

“The most intriguing aspect of this race is its potential,” The use of learned rules for robot control is a relatively new subject, and this 100-meter dash demonstrates superior performance compared to previous control methods.

I believe that progress will accelerate from this point forward.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Skip to content