A three-year-long challenge to find ways for robots to explore, navigate, map and do other tasks underground concludes with its final challenge this week.
The final event of the Defense Advanced Research Projects Agency Subterranean Challenge kicks off Tuesday at the Louisville Mega Cavern in Louisville, Kentucky, and teams of civilian researchers are competing to come up with the most useful systems.
The challenge matters to all in uniform whose missions may take them underground.
The tilt toward urban combat in megacities combined with the deep cave systems housing nuclear weapons in North Korea and other hazards in Eastern Europe means a subterranean combat scenario is likely for troops today and in the future.
In 2019, Military Times shadowed members of the 82nd Airborne Division as the Army began tackling subterranean challenges with soldiers. Things taken for granted above ground are either a major obstacle or impossible below the surface.
RELATED
Communications equipment isn’t always reliable, it’s dark basically all of the time unless you bring your own light and the confined spaces make for great ambush spots, dangerous gases can suffocate in seconds and there are always man-made dangers such booby traps.
Rather than send a living, breathing soldier or Marine down there, why not send a robot first?
All of the services have been working with ground robots for decades, often in tight, dark spaces.
But DARPA is trying to go one better — create autonomous or semi-autonomous teams of ground and aerial robots that work together to accomplish missions underground.
The team that wins won’t just get cash and bragging rights, but the chance to transition to a Department of Defense program that would see their work come to fruition and help troops.
The competition this week will put teams in an unknown, unmapped portion of the Mega Cavern both in real life and in a virtual version for other teams. The first time these teams will know what’s ahead of them is when their robot, real or virtual, rolls, walks or flies through the entrance.
Military Times spoke with the program head and members of one team each for the two different tracks of the competition.
Dr. Timothy Chung, the DARPA program manager for this challenge, said that over the course of the three-year competition, teams have faced urban-type settings such as sewers or tunnels. They’ve got to be able to work their robots on either a pedestrian walkway or a rugged, rocky cave footpath.
The teams will have an hour to retrieve and return 10 “artifacts” hidden around the cave system.
The artifacts include a thermal mannequin, which would put thermal cameras to use in rescuing a simulated survivor. Another is a cellphone that is hidden but transmitting, so a sensor that can detect the signal would help.
And as they’re doing their work, a kind of “dungeon master” can insert new problems such as jamming their communication network or creating a sudden virtual “rock fall” that blocks their path or cuts off one robot from the team, Chung said.
One is the systems track, that includes actual hardware, teams of robots running through challenges against other teams in a variety of underground settings. The other track is the virtual, in which competitors are working through virtual space, using lots of modeling and simulation to figure out what combination of robots or robot capabilities will work best in their competition.
There are 10 teams competing in the virtual track and eight teams competing in the systems track. Some teams are competing in both.
Systems competition prizes are: $2 million for first place, $1 million for second place and $500,000 for third place. Virtual competition prizes are $750,000, $500,000 and $250,000, respectively.
Professor Sean Humbert with the Multi-agent Autonomy with Radar-Based Localization for Exploration, or MARBLE team, told Military Times he’s worked previously on Army research for micro-autonomous teams doing intelligence surveillance and reconnaissance in urban environments.
Humbert’s team includes Professor Eric Few, Assistant Professor Chris Heckman and graduate research assistants Gene Rush and Harel Biggie.
The MARBLE team is also working a virtual challenge, and Humbert said that they’ve run a variety of simulations to work out problems for both competitions.
One of the biggest challenges that all of the teams face is communications. Radio frequency signals don’t travel well underground. And the key to the competition is having a robot team that coordinates its efforts.
Some of the teams have opted to drop as many beacons as they can, which act as a kind of router for the team, creating a robot-built comms network underground.
MARBLE went a different route, assuming there might be a jamming aspect, electronic warfare or just bad terrain that will prevent them from relying on their comms.
“Our team upfront realized that communications was going to be a really, really difficult aspect of the challenge here,” Humbert said.
In an earlier competition, their robot team was sharing a map to navigate, but when comms went down some robots didn’t have access to the map.
Their vehicles, both ground and air, won’t be communicating much with each other. They’ll disconnect from the network when they need to and run off on their own, following pre-programmed, autonomous functions.
They’ve got a small four-legged robot to help with difficult terrain and also four-wheeled vehicles that can work on a single level.
Each team has at least one aerial vehicle. MARBLE plans to use the onboard sensors on the small drone to help map the area. But they’ve only got about 10 minutes of flight time once it’s deployed from another ground robot.
The team halted any big adjustments to its algorithms and software about a month ago and have been stress-testing the robots ever since, smoothing out any wrinkles in how they should perform this week.
On the virtual side, players have a kind of “fantasy football” type of competition, Chung said. They are given a budget and allowed to mix and match the capabilities of virtual robots and even some robots in the systems competition that have been made virtual for that portion.
Dr. Sarah Kitchen with the Bayesian Adaptive Robot Control System Team has a mathematics background and is teamed up with experts in electrical engineering, computer science and computational physics.
Her team includes Reid Sawtell, Meryl Spencer and Richard Chase.
One of the advantages of the virtual option, she said, is the wider array of testing and options they can run through.
While they’ve got a lot to choose from, some basic characteristics stand out.
All of the underground environments are going to be interior spaces, feature challenging terrain, and have the possibility of a loss of visibility and a loss of communications, she said.
“Ultimately, it’s a 3D-connected space that may have some bigger and smaller pieces,” she said. “But in the topological space, it’s just connected tubes, essentially.”
The key aspect of their virtual team has been coordination.
They’ve got to set the virtual team in motion and step back, hoping their coding, software and algorithms — and plenty of mental gymnastics — have accounted for whatever the DARPA team will throw at them.
Todd South has written about crime, courts, government and the military for multiple publications since 2004 and was named a 2014 Pulitzer finalist for a co-written project on witness intimidation. Todd is a Marine veteran of the Iraq War.
