Skip to Content
MIT Technology Review

Why explosives detectors still can’t beat a dog’s nose

The oldest technology for detecting trace amounts of materials remains the best.

Golden retriever and artificial dog snoutGolden retriever and artificial dog snout
Golden retriever and artificial dog snout
Bob O'Connor

For nearly as long as armies have fought one another, they have enlisted animals to help. Horses, especially, were decisive for millennia. As historian Morris Rossabi has written of the Mongol conquest of Asia, “Mobility and surprise characterized the military expeditions led by Genghis Khan and his commanders, and the horse was crucial for such tactics and strategy. Horses could, without exaggeration, be referred to as the intercontinental ballistic missiles of the thirteenth century.” Historian David Edgerton notes that as late as the First World War, “Britain’s ability to exploit world horse markets was crucial to its military power.”

Horses are still of occasional importance, as in the American invasion of Afghanistan in 2001, when Special Forces troops on horseback called in bomb strikes via satellite radios, using laser designators and GPS reference points to guide the bombs. But horses are only very rarely the tool that separates defeat from victory: in all but the most exceptional circumstances, they have been replaced by tanks, trucks, satellites, and airplanes.


Yet while horses are largely gone from modern armies, dogs are not. As of 2016, the US military counted over 1,740 military working dogs among its ranks. At Lackland Air Force Base in San Antonio, the military breeds its own sleek puppies—mainly German shepherds and Belgian Malinois—who are groomed for military service from their first whimper. Some will wash out; others will go on to four to seven months of basic obedience instruction before receiving more specialized training in how to guard bases, ambush enemy combatants, and sniff out explosive devices. From there the field narrows further. The US Army estimates that to produce 100 war-ready dogs, it must train 200.

Before entering buildings in Afghanistan, Thomas, a US army paratrooper who asked to be identified by a pseudonym, would often send his platoon’s Belgian Malinois in first to ensure that no enemy soldiers or other surprises waited inside. During one day of particularly fierce fighting, Thomas was in a building, looking for somewhere to treat a wounded soldier, when he heard a noise from an adjacent room. As he rounded the corner to investigate, he remembers seeing “a shadow and a flash of light.” It was a Taliban-hired Chechen fighter with an AK assault rifle aimed directly at his face.

“You want to be able to sample the environment in a smart way, and dogs have given us a lot of insight into what that looks like.”

Just as the fighter squeezed the trigger of his weapon, the platoon’s dog came blazing into the room from the hallway and latched onto his neck, jerking him backwards. His shot was diverted, sparing Thomas’s life.

After that, Thomas brought the dog on every mission he could. “Sometimes people would say to me—‘Oh, you don’t need a dog for that,’” he says. “And I’d say, ‘Yeah, I need a dog. Are you on the ground? You’re not on the ground. I’m bringing the dog.’”

The military also relies heavily on dogs to sniff out explosives. Dogs’ sense of smell is estimated to be 10,000 to 100,000 times stronger than the average human’s. Billions of dollars’ worth of research on artificial detectors have yet to produce anything better. Unlike metal detectors, which are also used to locate roadside bombs and landmines, canines can be trained to pick up on non-metallic explosive devices concocted from fertilizer and other household items. This talent has proved particularly useful in Afghanistan, where many buried explosives are improvised from common chemicals packed into plastic jugs.

Scientists have long tried—and failed—to create devices capable of outperforming a dog’s snout. Starting in 1997, DARPA dedicated $25 million to an initiative called “Dog’s Nose,” which distributed grants to scientists to develop landmine detectors. At that point, an estimated 100 million mines were buried in approximately 60 countries. But according to the DARPA program’s director, Regina Dugan, the technology to find them had not advanced much since the Second World War. “The only landmine detection equipment issued to US soldiers in the field were the metal detector and a sharp, pointy stick,” she wrote in 2000. (The stick was to prod the ground for anomalies.)

The resulting machines, most of which featured polymer-coated tubes that reacted when exposed to explosives, seemed promising when used in sterile laboratories. But in more realistic environments things got messier. When one of the machines was pitted against landmine-detecting dogs at Auburn University in Alabama in 2001, the highest-performing canines were approximately 10 times more sensitive. In a 22-acre grassy facility in Missouri where DARPA invited participants to test their devices, some were too responsive, reacting to plants and soil in addition to explosives.

A decade later, in 2010, the commander of the Joint Improvised Explosive Device Defeat Organization (JIEDDO) admitted that despite a whopping $19 billion of government investment in spy drones, radio jammers, and aircraft-mounted sensors meant to combat improvised explosive devices (IEDs), dogs remained unparalleled as detectors of the dangerous devices. While sensors typically found half of the IEDs before they exploded, dog teams located 80% of them.

NIST

The newest artificial detectors can detect smaller traces of chemicals than a dog can. But those detectors are big, explains Matthew Staymates, a mechanical engineer and fluid dynamicist at the National Institute of Standards and Technology (NIST): “It’s got to plug into a wall, you need an enormous amount of infrastructure, gases, and vacuum pumps—and you have to bring the sample to your machine.”

Nonetheless, artificial detectors have a role to play in places like airports, where all passengers must pass through security checkpoints, and dogs have provided inspiration for improving them. Staymates used a 3D printer to replicate the nose of a female Labrador retriever named Bubbles. The result is a snout-shaped extension that goes on the front of commercially available explosives detectors. It sniffs air like a dog, inhaling and exhaling several times a second instead of continuously sucking air in as such machines normally do.

The researchers found that this method, counterintuitively, pulls in samples of air from farther away, drawing in more of the chemicals floating around. “Nine times out of 10, you don’t know where the bad guy with a pipe bomb in his backpack is,” Staymates explains. “So you want to be able to sample the environment in a smart way, and dogs have given us a lot of insight into what that looks like.”


Despite this progress, a dog is still much more effective than an electronic bomb-sniffer—not least because an animal, like a human but unlike a machine, can react to unpredictable situations. So some scientists have focused their efforts not on replacing working animals, but on improving their performance.

In 2017, a team at MIT’s Lincoln Laboratory developed a new mass spectrometer, about the size of a large dresser, that could identify trace amounts of chemicals on a par with canine performance. Not only was it impressively sensitive, but it was fast, completing its assessments in about one second. The researchers were excited about the device’s potential not to substitute for bomb-sniffing dogs, but rather to help train them.

The team had dogs locate explosives previously hidden in canisters, which were also analyzed with the spectrometer. The machine discovered that some of the perceived errors the dogs made—identifying explosives in supposedly empty vessels—weren’t errors at all; the containers had been cross-contaminated. That allowed the trainers to better regulate when to praise and reward their canine students, reinforcing their detection abilities.

Though some labs wanted to adapt the machine to replace dogs, the MIT team disagreed. In a news release at the time, Roderick Kunz, who led the research, said: “Our feeling is that such a tool is better directed at improving the already best detectors in the world—canines.”


Haley Cohen Gilliland is a writer in Los Angeles.