PITTSBURGH – Starting tomorrow, autonomous robots from 11 teams will descend into an underground mine outside of the city for the first stage of DARPA’s multi-year Subterranean Challenge. Teams will face a simulated mine disaster, with a mission to create 3D maps of the underground mine, and identify several prepositioned objects and “human survivors.” With a combination of robots and drones to navigate the environment, teams will have 60 minutes each to complete the mapping and search mission.
This is the latest in a series of DARPA challenges, held since 2004, that focuses on pushing the envelope of what is technologically possible. Well-known past DARPA Challenges include the DARPA Grand Challenges (2004-2005), which focused on autonomous vehicles, and the DARPA Robotics Challenge (2013-2015), which focused on semi-autonomous robots in disaster response situations.
“DARPA always tries to pick a problem that is close enough to the edge so that people can potentially tackle it, but far enough away that it’s not a done deal,” said Joel Burdick, professor of mechanical engineering at the California Institute of Technology (Caltech), and leader of Caltech’s contribution to NASA’s Jet Propulsion Laboratory’s (JPL) team.
Burdick said he’s one of only two or three people in the world to be involved in each of DARPA’s challenges, and “we’ve learned things in all of the previous challenges that have actually opened up new research directions for the people involved.” For example, many of the people who are now players in the autonomous car industry participated as graduate students in DARPA’s autonomous car challenges.
“There have been a lot of demos running robots in mines and creating maps over the years,” said Sebastian Scherer, associate research professor at Carnegie Mellon University’s Robotics Institute, and co-lead of the CMU/Oregon State team. “But the challenge that DARPA is pushing, is how to create a system that is robust and resilient. There are so many things that can go wrong in an exploration where you don’t know what’s ahead of you.”
DARPA has given few clues to the teams about what the course will look like. “There’s a big mobility aspect to this,” said C.J. Taylor, professor of Computer and Information Science at the University of Pennsylvania, and the leader of Penn’s team. “These underground environments are difficult, dark, damp, and cold, and have uneven terrain that presents all kinds of hazards.” Sensing is an important part of this challenge, and the competitors are tasked with identifying various artifacts along the way. Mapping is also critical, and each team will need to drop sensors around the mine to map the terrain. Then, “there’s an added problem of coordination in an environment where communication is limited,” continues Taylor, since radio waves don’t penetrate well underground. “That’s sort of what excites me about the challenge – the fact that it requires the team to stretch in so many directions.”
DARPA has also been good at funding organizations for its challenges that have various approaches to the problems it is trying to solve, notes Burdick. “The design space is so large, and different teams have different strengths.” There are some participants that are leaders in drone technology, while others have expertise in ground robotics. “JPL is known as a system integrator,” said Burdick, “so that’s a lot of what we’re focusing on: how to build the entire system and not shortchange any individual aspect of it.”
Each team has made different choices about what robotic technology to deploy and the way they address communication and control. “Hopefully each of us can contribute something to push the state of the art forward,” said Taylor. Burdick agreed: “Previous challenges have been fantastic geek fests where you show up and see what all the other teams are doing.”
“DARPA has established an all-star cast here,” Taylor added. “We’re really impressed by the kind of talent they’ve been able to bring to bear in the challenge.”
While applications for law enforcement, the military, and first responders come easily to mind, there are other applications for this kind of technology as well. Astrobiologists and astro-geologists at JPL and NASA, for example, have been thinking a lot about how to map caves on Mars or lava tubes on the moon. Maintaining communication between the robots and earth will obviously be difficult, so autonomy like that being tested in the Subterranean Challenge will be critical.
While all the teams are ultimately trying to win the $2 million prize to be awarded next year, Scherer from CMU sums up the view of a lot of the participants. “If we can deploy all the robots in the mine and make it back out with maps and get some points, that would be a huge success.”
Editor’s note: A media day for the challenge is scheduled for Wednesday, Aug. 21, to give updates on the challenge. Robotics Business Review plans to have additional coverage of the DARPA Subterranean Challenge following this event.