- By Alexis Madrigal
Semi-autonomous robots that can navigate and map drug-smuggling tunnels could be the greatest weapon to emerge from the government’s attempt to stamp out the trade in illicit substances across its borders.
Using special intelligence software developed at Idaho National Laboratory that can be mounted on different machines, the iRobot and Foster Miller robots use lasers to situate themselves in the dark tunnels that have been bored beneath the line that divides Mexico from the United States.
The subterranean passageways are a tough environment for Border Patrol to police. The agents know nothing beyond that there’s a hole in the ground. Some tunnels turn out to be crude holes. Others can reach three-quarters of a mile long and be part of a complex distribution infrastructure.
“They are not places you want to send people, especially ones that are claustrophobic, so it’s a perfect application for robotics,” said INL roboticist David Bruemmer, who has spent a decade developing the software the robots run. “That’s where we’ve really found a niche for the capabilities that we have.”
High-tech border surveillance has taken off since both the Sept. 11 attacks and the surge in illegal immigrants over the last decade. Tech is playing a bigger and bigger part in Border Patrol efforts because it’s simply too expensive to have agents everywhere. But for every high-tech solution, say, the controversial “virtual fence” that will begin construction soon in Arizona, there’s a low-tech countermove: mole tech.
In Arizona, more than 30 tunnels have been discovered just since 2006, when Congress passed the Secure Border Fence Act, which called for the construction of more than 700 miles of fence across California, Arizona, New Mexico and Texas.
Near San Diego, 32 tunnels have been discovered since Sept. 11, 2001. Before the heightened vigilance that came in the aftermath of the attack, only two tunnels had been discovered in eight years. Now, there’s a special multi-agency task force in the area dedicated to stopping the tunneling operations. In the past, officers have found them more-or-less by chance.
“We’ve discovered people coming out of the ground on camera footage,” said Jerry Conlin, a Border Patrol agent and spokesman for San Diego. “We’ve had others where we had an agent witness someone disappearing into the ground.”
At the end of 2007, a canine unit picked up a scent and followed it to a storage facility near Tecate, Mexico. Agents eventually pulled 13,700 pounds of marijuana out of the tunnel they found on the premises.
“With our increased operational control, it has literally forced them to go underground,” Conlin maintained. “We disrupted the traditional smuggling routes.”
That’s created a new need for ways of fighting the subterranean drug trade. Sophisticated geological techniques for detecting tunnels offer one solution, but once you’ve found a tunnel, you’ve got to figure out what’s down there. That’s where Bruemmer’s bots — running what his lab calls the Robotic Intelligence Kernel — come into play.
In December, they brought a sensor-loaded Talon robot to a tunnel that the Department of Homeland Security had seized. Though it had been entered, the government agents knew little about the space. Victor Walker, another roboticist at INL, accompanied the bot to the border near Arizona.
“They brought us to this warehouse,” Walker said. “There was a grate in the ground around back. Just a drain. I did not expect that. The Talon is a pretty big robot. You pulled it up and it dropped down about 10 feet below the warehouse.”
This anteroom to the tunnel was dark and damp, and about the size of a large bedroom. In the corner, was a shaft that dropped fifty feet down to the tunnel proper, which ran about 90 meters. They lowered the robot down with wires and, after a few technical hiccups, traversed the muddy hole. It output chemical readings, video, and a map like the one you see below, which can be stuck into Google Maps.
“We hooked it up with a chemical sensor. We were able to map those chemicals to the map. You could see as it was going along,” Walker said. “Within a few minutes, we were able to task it down and get the video back so [Homeland Security officials] could look at it.”
A map generated by a robot with software developed by Regis Vincent at Stanford Research Institute.
The INL robots aren’t the only ones being used by government officials, nor is the border the only place where robotic border inspectors might be used. Canadian robot maker Inuktun specializes in pipe inspecting robots operated by human beings. Most of the more than 1,000 bots they’ve sold are used by utilities checking out their sewer pipes or water mains. In recent years, however, they’ve seen requests from government agencies to repurpose their bots for subterranean inspection.
“I don’t think anyone has ever built a robot to go into a tunnel, but if you’ve built a bot to go into a nasty sewer pipe, it translates fairly well to going into a tunnel,” Dobell said. “There are a lot more of these cross border tunnels than people think.”
The company’s president, Colin Dobell, said that he could not reveal the names of the organizations that he’s working for, but that he knew they’d been deployed.
“I can tell you that they have been used in tunnels and have been used in tunnels that go across borders,” Dobell said.
The INL bots, though, use a fundamentally different control paradigm. Dobell’s bots are teleoperated, meaning there’s a human with a joystick driving them around. Bruemmer’s are a kind of hybrid bot that share control between the operator and the robot. Operators tell Bruemmer’s robots where to go, but the robots drive.
In the tunnel application, the robots use their lasers to locate themselves within the space and help human operators controlling them with a standard joystick or a Wiimote from running them into walls.
“In the Arizona tunnel, there was less than an inch involved on each side,” Bruemmer said. A teleoperator without some guided motion couldn’t do it.”
Most importantly, though, they can go exploring and mapping autonomously. Inside tunnels, you can’t always communicate via the standard means with the robots. If they go deep and far enough away from the operator, they’ll lose communications contact. In that case, the operator can set a time limit for autonomous exploration, which, when it expires, will send the robot back into communications range to phone home the data it’s found. That information is integrated into the operator’s heads-up display, and then the robot can be sent a-roving again.
Robots might not be serving us drinks yet, but they are evolving to suit our real needs. And slowly but surely we’re learning how to take advantage of robots’ potential.
“It’s all about the man-machine interface. Like Windows just provided this simple user-understood interface, I think that’s what we’re really trying to do with robots,” he said. “Forget about trying to make robots massively intelligent.”