- Dangerous chemicals may be an industrial necessity, but accidents can leak them into the environment
- Tracking the plume of airborne pollutants can help move people out of harm’s way
- Chemical-sensing drones provide mobility and more robust pollutant models
When imagining major weather catastrophes, most concerned citizens think about homes and businesses damaged by gale-force winds. However, many people don’t realize that immediate physical destruction is often only the beginning.
Take the case of Arkema North America, which had a chemical plant in Houston, Texas, during Hurricane Harvey. A combination of the storm’s devastating power and corporate negligence caused a chemical leak that forced 200 people out of their homes and required 21 to seek medical attention.
Accidents happen during major events like a hurricane, but hazardous materials becoming airborne are especially dangerous because they’re difficult to track. For Dr. Edward Knightly of Rice University, the answer is simple – chemical-sensing drones.
Wind patterns are tough to predict after a major weather event, but the problem is compounded when airborne chemicals are added to the mix. Knightly explains that this variability makes tracking volatile organic compounds (VOCs) extremely difficult.
“A compelling example was last year during Hurricane Harvey when some of the chemical plants flooded in the shipping channel and there were fires,” says Knightly.
“Fires at petrochemical plants were releasing chemicals and it was not known what the evacuation area should be. In fact, first responders tried to evacuate a perimeter around one of the fires and the first responders themselves ended up hospitalized.”
While many crises still require humans to risk their lives, this is a scenario where an autonomous vehicle could perform the necessary task. As Knightly explains, these specifically-programmed drones are able to create a 3D map of a chemical plume.
“When we’re out sensing in the field, we’ll have a mixture of different gases, so we train in the lab to evaluate the signatures for different compounds so that in the field we can identify the parts per billion concentrations of each compound,” says Knightly.
“The drones will not only sense, but also learn about the mix of compounds in the environment in order to find out what are the dominant chemicals signatures to find the emission sources and identify safe border areas.”
This machine learning process is meant to allow the drones to operate with little human interference. An accompanying mobile app called ASTRO will help alert local residents to hazardous VOCs in the air.
“We’re still designing the sensor, but the core technique is based on laser spectroscopy. My collaborator, Frank Tittel at Rice, is one of the world experts in that area,” says Knightly.
“The basic idea is that when you shine a laser through a gas, the different components of the gas have different spectral signatures. When there’s a chemical in the air, whether its methane or benzene, there’s a big spike in the spectral signature. We can potentially, with different lasers and different frequencies, be able to detect any chemical. We’ll be starting with some of the simpler ones as baselines and then moving on to more complex signatures.”
More money, more research
This research could go on to save countless lives, and it’s caught the attention of the National Science Foundation (NSF) which has awarded Knightly and his collaborators $1.5 million to further expand ASTRO.
But just because this is serious science doesn’t mean that Knightly’s team isn’t having a little fun.
“Right now we’re doing a lot of testing on Rice’s campus,” says Knightly. “Because we don’t have hazardous gases here, we’re doing early experiments with wireless signatures.”
“One of our early experimental milestones will be to ask the drones to find a designated wireless device. We typically put the wireless transmitter in a bicycle and somebody rides the bicycle around campus, and then the drones have to find it and track it.”
Knightly believes that this shift toward autonomous behavior is the next big step for drones. If scientists can learn to harness technologies like machine learning, drones will be able to learn anything from search-and-rescue to package delivery.
“I think we’re at the very beginning of seeing what drones can do,” says Knightly.