Atlas, The Famous Humanoid Parkour-Robot, Is Back Like A 10-Point Gymnast

The bipedal humanoid can perform new routines with a success rate of 80 percent. Boston Dynamics/Screengrab

Atlas, the famously agile humanoid, is at it again. This time, sticking a parkour-style gymnastics routine worthy of a gold medal. 

You may recall two years ago when Atlas seriously struggled to place a box on a shelf, a hapless display serving as a metaphor for every Monday ever.

Last November, Atlas made headlines again when it showed off its super savvy parkour skills, jumping from block to block and even throwing up a serious back-flip

Atlas is back again in 2019, this time as a fluid 10-point gymnast, maneuvering with ease and grace as it handstands, tucks, and somersaults its way to glory. (That little split-kick "wee!" at the end though.)

Atlas is described by its creator Boston Dynamics as the "world's most dynamic humanoid robot" and has pushed the robotics research platform since its first introduction to the world in 2013. Originally measuring nearly 2 meters tall (more than 6 feet) and weighing almost 150 kilograms (330 pounds), Atlas was developed for the US Defense Advanced Research Projects Agency (DARPA) for use in the agency's Robotics Challenge. At the time, Atlas was one of the most advanced humanoid robots ever built. Now nearly a decade on, the bipedal powerhouse continues to tickle our uncanny valley fantasies. 

These days, Atlas is made of 3D-printed parts weighing just 80 kilograms (176 pounds) and with a "strength-to-weight ratio" that allows it to leap and somersault like an Olympian. With a speed of 1.5 meters (5 feet) per second, Atlas is also dynamic AF. 

"Atlas’s advanced control system enables highly diverse and agile locomotion, while algorithms reason through complex dynamic interactions involving the whole body and environment to plan movements," writes Boston Dynamics on its website. The sure-footed robot is made up of "custom motos, valves, and a compact hydraulic unit" touting 28 hydraulic joints for "impressive feats of mobility." 

In its new rendition of mobility maneuvers, the creators of Atlas first used an optimization algorithm to transform high-level descriptions of movements into motions that were "dynamically-feasible" for the robot. Atlas then tracked the motion using a model predictive controller that allowed it to seamlessly blend each movement into the next. Boston Dynamics says that this approach allowed them to develop the routine "significantly faster" than previous ones with a success rate of about 80 percent. 
Comments

If you liked this story, you'll love these

This website uses cookies

This website uses cookies to improve user experience. By continuing to use our website you consent to all cookies in accordance with our cookie policy.