As It Happens6:14Cassie the robot runs a record-breaking, 100-metre sprint
If you’ve got not too long ago run monitor at Oregon State University, chances are high that you’ve got crossed paths with an uncommon competitor: an athletic, bipedal robot.
Cassie first made a identify for itself in 2021 by finishing a 5K on campus with a time of 53 minutes. After months of coaching, the torso-less, faceless machine has now set a Guinness World Record for the quickest 100-metre sprint by a bipedal robot.
It clocked in at 24.73 seconds, which is still removed from Usain Bolt’s record-breaking human time of 9.58 seconds.
“Cassie runs it like a average jog,” engineering pupil Devin Crowley advised As It Happens host Nil Köksal. “It’s spectacular for a robot, but it’s not fairly harmful but.”
WATCH | Cassie runs a world record:
After finishing a 5K final yr, Cassie units a Guinness World Record for the quickest 100 metres by a bipedal robot. (Dynamic Robotics Laboratory/Oregon State University)
Crowley spent six months coaching Cassie for the 100-metre sprint, which occurred on May 11. He’s happy with the robot, though the sight of it sprinting throughout OSU’s Whyte Track and Field Center was otherworldly.
“I like to explain it because the decrease half of the Terminator,” he mentioned. “But it’s extra like an ostrich…. It has inverted knees. Technically, they’re … not knees; they’re ankles. So it seems to be like a chicken. Just a pelvis with legs, and so they bend principally backwards.”
How to construct a operating robot
Oregon State robotics professor Jonathan Hurst, synthetic intelligence professor Alan Fern and their college students developed Cassie’s construction and software program in 2017. They used funds from the National Science Foundation and a $1-million US grant from the Defense Advanced Research Projects Agency. They additionally lead the robot to its 5K feat.
Crowley constructed on that work throughout his remaining grad research mission to coach Cassie for the sprint.
The robot needed to begin by standing still, after which speed up and run till it crossed the end line. It then needed to decelerate and return to a standing place — which was the most important problem for Crowley. He says he specialised Cassie’s programming to run as quick as attainable, by prioritizing velocity over different attributes, akin to turning.
University of Bristol robotics skilled Nathan Lepora known as Cassie’s sprint “vastly spectacular.” He as soon as helped develop a robot that may gently shake arms utilizing a man-made fingertip with a sense of contact.
“I seen [Cassie] stumbled a few instances,” he wrote in an e mail to As It Happens. “Humans depend on our sense of contact in our toes to assist us run — you’ll be able to’t stroll very properly with a numb leg. So possibly these robots could possibly be even higher if they’d tactile toes, then they may be capable to run quicker and even over uneven floor.”
He mentioned it’s nice to see robots like Cassie coming from a college analysis lab, somewhat than a non-public firm like Boston Dynamics.
“Much of the analysis on this space … has been achieved privately, so different lecturers haven’t been capable of construct upon the analysis,” he mentioned.
“Legged robotics is a vital space as a result of it permits robots to maneuver freely round human environments. It is tough to go up stairs if you’re on wheels.”
An evolutionary bot
Crowley had a lot of enjoyable coaching Cassie for the 100-metre sprint. He says he used a learning-based strategy to high-quality tune the robot into a runner.
“We do not inform it easy methods to run. We inform it when it’s doing a good job at operating,” he mentioned. “It performs round. It tries random actions, primarily, till it found out which issues we inform it are finest. And as a results of this strategy, it’s rather more sturdy.”
The coach in him referenced some moments within the video of the record-breaking run, when Cassie was about to wipe out, but was capable of recuperate.
“That’s actually cool to see,” Crowley mentioned.
But Cassie has still been bumped and bruised over time — and its bodily physique will quickly retire.
The OSU Dynamic Robotics and AI Lab has entry to a new model, to which they may add Cassie’s software program. The subsequent model of Cassie can have imaginative and prescient in order that it’s no longer operating blind, actually.
“There are a variety of completely different instructions that Cassie’s going,” Crowley mentioned.