Watch a gallery of the robots in action
Inside Building 9, a concrete hangar in eastern Los Angeles, the pressure is mounting.
There is a hum of activity as engineers ready their hyperadvanced robots for the final run in the world’s most important robotics competition: the, which ran over Friday and Saturday last week at this former horse-racing track in Pomona, California.
A whiff of dung hangs over the course as robots scramble over debris, drive emergency vehicles, climb stairs and cut through plasterboard walls – tasks designed to simulate those. There are 23 competitors, four at a time on each task, which they attempt twice over the two days.
A human would whiz through the course in minutes or seconds, but only the most advanced machines can tackle it, and then only with human controllers. But the data they generate is forming the bedrock of robotic autonomy.
Teams of roboticists have. I join the few thousand people who have turned out to watch how well these machines can
Gulf in performance
I instantly see there’s a gulf in performance between the top robots and all the others. A handful of bots can’t even get out of the vehicle they drive to the start of the course. At one point,spills head-first out of its vehicle, tearing its hydraulic lines and spraying fluid in an arc.
Some machines, such as the spider-like Grit from a startup in Colorado, never make it off the starting line.
The action is interspersed with long pauses (some spectators compare it to golf) as the robots survey the environment and calculate their next steps. Some, like MIT’s humanoid Helios and CHIMP from Carnegie Mellon University, carry out many steps automatically. Others, like the wheeled rover Momaro from the University of Bonn in Germany, are entirely human-controlled.
“The human operator is as much a part of the system as the robot,” says Brett Kennedy, who leads the team from NASA’s Jet Propulsion Laboratory with the squat, four-legged robot
By the end of the first day, the teams have begun to hone their skills. Momaro zooms its car down the course and gets out in a flash using its extendable legs, each with a large wheel on the end.
Robosimian drives slowly and carefully, then climbs out of the vehicle like a monkey, placing the equivalent of its hands on the ground to steady itself. It then folds into a squat and starts to roll forward on wheels, a much easier option than coordinating its spider legs. The robot progresses slowly but smoothly, without any mistakes.
Momaro steams ahead. It is by far the fastest robot, and makes almost no mistakes. It and Robosimian jockey for the lead on their run and eventually both collect 7 points out of 8 in less than 50 minutes – high scores and good times, though neither tackles the stairs. The other robots on their run are impressive too. Although Helios falls twice and has to be winched back to its feet, it is fastest of all at climbing stairs.
CHIMP and its Tartan Rescue team scored well in a different area. It’s the heaviest robot, a big red beast that runs on treads, weighing 200 kilograms and able to lift 136 kg with just its arms. In one run, CHIMP neglects to turn on its drill before trying to cut through a wall, but is still able to tear through it with sheer strength.
CHIMP finishes by powering up the stairs, using its treads to roll up, rather than stepping. This helps it top the rankings after the first day, the only team to score the full 8 points.
While the crowd is cheering for the robots outside, the heart of the operation is back in Building 9. A 5-minute walk from the course, it’s where each team runs its control centre, guiding their robots over a wireless link, without cables for power or data, something rarely done before this competition.
The mesh between machine and operator is crucial, and each team approaches it in a different way. Some, like Florida-based IHMC, rely on a single master controller – the teams call them robot whisperers.
John Carff, IHMC’s pilot, and a video game whiz, is widely considered to be the top robot whisperer. Carff controls Atlas – also known as Running Man – almost entirely by himself, manoeuvring it around a virtual replica of the course, built using data from its sensors.
Others rely on a group of people: Momaro has nine, for example. Sebastian Schuller controls the arms and uses an Oculus Rift to peer into a virtual world in true 3D generated by Momaro’s on-board laser scanners. “To interact accurately with the environment, I need to see from the perspective of the robot,” Schuller says. “At first it’s a little bit trippy, but you get used to it.”
If you would like to reuse any content from New Scientist, either in print or online, pleasedepartment first for permission. New Scientist does not own rights to photos, but there are a available for use of articles and graphics we own the copyright to.
This entry passed through the Full-Text RSS service – if this is your content and you’re reading it on someone else’s site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.