Cassie the bipedal robot runs a 5K – TechCrunch



Chances are you’ll properly acknowledge Cassie as the premise of Agility Robotics’ supply bot, Digit. In case you’ve been following the tech’s development in any respect, nevertheless, you little doubt know that it began life because the ostrich-inspired Cassie. The robotic is all legs and never a lot else.

Along with fueling Agility’s business ambitions, the robotic has confirmed a stable platform for exploring bipedal location. Introduced by Oregon State College professor Jonathan Hunt in 2017, Cassie was created with the goal of a $1 million grant from the DoD — a reasonably acquainted story within the robotics world.

As we speak, a staff from the Dynamics Robotics Laboratory in OSU’s Faculty of Engineering highlighted the methods they’re persevering with to push Cassie to its bipedal limits. The staff says the robotic was in a position to run a 5K untethered, on a single cost. Cassie’s not going to beat any human world information anytime quickly, however the 53-minute (and three second) run was nonetheless a formidable exhibition for the tech.

The robotic’s run time included round 6.5 minutes of troubleshooting, because the staff handled an overheated laptop and a botched flip that knocked it off its legs.


“Cassie is a really environment friendly robotic due to the way it has been designed and constructed, and we have been actually in a position to attain the bounds of the {hardware} and present what it could do,” Ph.D. pupil Jeremy Dao stated in a press release.

In keeping with the staff, Cassie basically taught itself to run utilizing a deep reinforcement studying algorithm, which allowed the system to determine how you can keep upright by shifting its stability whereas operating.

“Deep reinforcement studying is a strong methodology in AI that opens up expertise like operating, skipping and strolling up and down stairs,” undergrad pupil Yesh Godse provides.

In Might of this 12 months, the OSU staff additionally demonstrated how Cassie can stroll up and down stairs with out the help of lidar or on-board cameras.


Supply hyperlink