In the future, farmers will be able to look at a tablet and see the nutrient, CO2 and moisture levels of every individual acre in their fields in real time. This will allow them to irrigate and fertilize more strategically, reducing waste and increasing yields. But while there are all sorts of sensors to measure soil health, the long-range, low-power, cost-efficient wireless networks needed to transmit that data from the back 40 to the barn don’t exist yet.
Bhuvana Krishnaswamy, an assistant professor of electrical and computer engineering at the University of Wisconsin-Madison, hopes that her work optimizing the algorithms used by such networks will help plant the seeds of a new agricultural revolution. Her research is supported by a five-year, $500,000 National Science Foundation CAREER Award.
“Right now, most farmers use one sensor to make decisions about their whole field. We want to deploy these sensors and collect data over long distances so we might be able to map out a farm and make decisions about irrigation and fertilizer distributions,” she says. “There’s a lot of heterogeneity across a field. This could provide farmers with more data.”
When Krishnaswamy attended a welcome event after joining UW-Madison in 2018, she found herself chatting with a soil scientist. They wanted to pick her brain about any available commercial wireless sensing technology that could cover a kilometer or more. Krishnaswamy didn’t know of any off the top of her head, and when she looked further into the matter, she found that the few long-distance sensor networks out there were prohibitively expensive.
“Like everyone else, I wondered why they couldn’t just use cellular networks, satellite networks, or Wi-Fi?” says Krishnaswamy.
It turns out, Wi-Fi is not optimized to cover dozens or hundreds of acres of remote land. And using cellular networks is prohibitively expensive, since each sensor needs its own SIM card with the associated data costs, and the same goes with satellite connections. “From what I hear from my collaborators, a one-cent increase per pound of product is still too much for farmers, because their margins are pretty low,” says Krishnaswamy. “So we want to keep the cost as low as we can.”
To create a cheap, reliable network that can handle data from dozens or hundreds of sensors, Krishnaswamy is designing algorithms both for the transmitting sensors and the receiving station that leverage the fact that each sensor only sends a tiny amount of data every 30 minutes or so. By using signal processing techniques, she will enable the receiver to listen to, and decode multiple signals in parallel. The project will also involve cloud computation that will use machine learning to fix any data lost in the transmissions. Altogether, these changes to the network will allow a large number of sensors to transmit tiny packets of data without draining their batteries.
While her CAREER proposal focuses on the algorithms, Krishnaswamy says the hardware side of the project is already moving ahead. Her team and her collaborators in soil science are currently building a “black box” that includes an array of off-the-shelf sensors that can collect data on moisture, CO2 levels, pressure, temperature and other variables that they will field test in summer and fall 2022. Eventually, the researchers hope they will be able to test a network of dozens or hundreds of sensors running Krishnaswamy’s algorithms.
The networks have potential applications outside conventional farming. Other researchers have shown interest in using the long-distance, low-power networks to communicate with solar panels as part of a distributed power grid. Others are interested in using them to model microclimates in vineyards or in cities to create a large network of pollution sensors.
Krishnaswamy hopes to get rural communities involved in the work as well. Through connections with the UW Extension, she and her team hope to enroll Wisconsin farmers to field-test their sensor arrays. She also hopes to host a rural science teacher for a summer in her lab to learn more about the sensing technology. “Many high school students that take agriculture classes choose a career in farming,” she says. “We’re hoping our teacher will be able to expose them to programming, simple hardware designs, and help develop skillsets in designing circuits as a way for students to understand how to use sensing and computer technologies to improve farming practices.”