There are robots that can see. There are robots that can hear. But there aren’t many robots that can smell. Dr. Lorena Villarreal, an assistant professor from the Department of Electrical and Computer Engineering, is changing that with her innovative robotic nose, an award-winning invention she hopes to see deployed in real-life disaster scenarios in three years’ time.
The device landed her on the MIT Technology Review’s “Innovators under 35 Mexico” list in 2014, and she has spent the intervening years continuing to perfect and refine her design with some extra help from DigiPen students.
Villarreal grew up in Monterrey, Mexico, and was intrigued by electronics at a very young age. “My father is an engineer, so I used to see him building and fixing things after work, things like the VCR, television, remote control toys, everything we as children broke,” she laughs. Inspired, she began assisting her dad on household repairs, learning enough by age 13 to fix things like a broken ventilator and a curling iron by herself. When she and a group of fellow women tried to join their middle school’s electronics club, they were banned from doing so by the school administration.
“Women weren’t allowed,” she says. “Women could only take cooking, drawing, sewing, or learn to be a secretary. I tried to gather signatures and stuff like that to make them teach us, but it didn’t work.” Eventually, that mentality did change though, which Villarreal got to see from a unique perspective. “Two years ago I was able to see my middle school winning a robotics regional competition where three of the five students on the team were women. I was a judge in that competition for other categories,” she says.
Villarreal had to wait until college to finally pursue her passion in an academic setting, studying mechatronics (a combination of mechanics, electronics, control systems, and programming) and eventually earning her Ph.D. in robotics and intelligent systems. It was during her post-graduate studies that the idea for a robotic nose occurred to her.
“There had been some disasters in Mexico at that time, hurricanes mostly, and I found out that one of the university departments was doing research about rescue robots,” Villarreal says. “I wanted to make mobile robots, so the idea just came out while doing research about what was needed.”
The field, unsurprisingly, was full of research on robotic vision and sound, but research on implementing smell in robots was scarce. Sensors that can “smell” did exist, but their implementation was largely limited to the chemical industry. “For example, food quality assurance,” Villarreal says, “the sensors can read differing quantities of chemicals in a sample, so with the proper technology it can say, ‘Oh, this is a soda, and it’s a Pepsi.’”
In a rescue situation, a robotic nose with a chemical smelling sensor could be incredibly useful for tracking and detecting the source of dangerous gas leaks.
When it came to design time, Villarreal studied the masters.
Using a special slit, dogs can inhale and exhale in different directions. This makes them incredibly efficient trackers.“I looked at human noses but also bear noses, which can smell very long distances, and dog noses,” Villarreal says. “In terms of tracking, the dog nose was the most helpful. If you’ve ever looked closely at a dog nose, you’ll notice that they have a septum that’s very different than ours.”
“When they are on the floor sniffing, what they’re doing is inhaling the odor while quickly expulsing the air without affecting the trail of odor chemicals they’re trying to track,” Villarreal says.
In designing her robotic nose, not only did she mimic nostrils and the structure of a dog’s septum, she also gave the robot the ability to inhale and exhale like a real nose through the use of a ventilator, which improves the sampling process. Given that the robot is meant to detect dangerous gas, all the air exhaled by the robot passes through cleaning filters. Last year, her patent for this design was granted in the U.S.A. and Mexico.
While her first designs were scrappy, utilizing bottles and cartons in combination with chemical smell sensors, they were also highly positive proofs of concept. “I placed two sensors, not in nostrils at that point, but just two isolated sensors, and it could figure out the position and direction of the smell, so I knew that it would work,” Villarreal says.
Today, her robotic nose design is much further along, but there’s still work to do. Before the robot makes its way out into real disaster scenarios, Villarreal has specific improvements she wants to make. “I want to build a more robust version of the robot,” she says, one that can weather dust and debris-filled buildings. The other is her continual quest to make her design more efficient. DigiPen students have been helping her do just that over the summer. By running 3D airflow models, the team discovered they could improve the design by changing the sensors’ position, in turn allowing Villarreal to reduce the nose’s overall size.
Her newer models have also incorporated a camera with the objective of identifying pipes, allowing the robot to rely on sight as well as smell to track gas leaks. The other key ingredient in making the robot more efficient lies in the algorithm that tells it when to turn left, right, or take a moment to sniff, which Villarreal is once again improving by mimicking biology, this time through artificial intelligence.
“I’m using evolutionary algorithms, which means an AI creates random different algorithms, picks one, runs it through experiments, picks the next one, then runs it again and so on until all the algorithms have been tested. Then they ‘evolve’ when the AI selects the best ones and mixes them,” Villarreal says. “It’s ‘evolutionary’ because it runs this process many times and theoretically, by the end, the best algorithm survives.”
In the meantime, Villarreal has also started imagining other commercial uses for her robotic nose, most of which she wants to keep under wraps for now. She did let us in on one intriguing idea though, one that relates to a lesser type of disaster. “Say you have a nose on a cleaning vacuum robot in your home. Instead of just going everywhere, it could smell for something that’s spilled. It could be anything, and go, ‘Oh, something is here. Let me clean that,’ and go straight to it instead of covering the whole place, draining the battery and wasting resources. … We still have a lot of research to do.”