Liu’s “IoTree” sensor is the next step in smart agriculture

As anyone with a houseplant or a garden knows, it can be surprisingly difficult to determine how much water and fertilizer a plant needs. Finding a way to accurately tell what a plant requires at any point during its growing cycle would offer incredible benefits to agriculture, horticulture, and forestry.

“Developing this type of sensor would have many impacts,” said Jian Liu, an assistant professor in the Min H. Kao Department of Electrical Engineering and Computer Science (EECS). “We could track pollution levels, climate change, or rate of habitat loss; we could leverage plant health indicators to guide pest control strategies or use nutrient level data to inform fertilizer application and timing.”

Networks of interconnected devices, known as the Internet of Things (IoT), are gaining popularity to monitor and manage complex systems. Some monitoring networks have even been deployed for agricultural plants, but they are far from ideal.

“Existing plant sensor systems are very bulky and battery-powered, so you are not able to scale their deployment easily,” Liu said, explaining that outfitting a field of just 100 trees can be prohibitively expensive.

A typical apple orchard in the southeast United States of America can have 10 times that many trees in a single acre.

Liu is stepping up to make plant health data easier to collect and more cost-effective. Last August, he received $300,000 to develop software for a plant monitoring system that will deliver unprecedented levels of data about a plant’s internal state. The funding is part of a $600,000 National Science Foundation (NSF) grant headed by VP Nguyen, an assistant professor at the University of Massachusetts Amherst (UMass Amherst).

“We are trying to build a low-cost, battery-free sensor that you can deploy and scale very easily, so you can monitor the status of trees in a very large area,” Liu said.

Designing Software to Last a Lifetime

Over the next three years, Liu and Nguyen will develop a sustainable, zero-maintenance, biocompatible sensor system that trees ‘wear’ throughout their lifetimes. Each device in the system, which the team calls IoTree, will sit mostly on the outside of the tree trunk, with a thin multi-chemical sensor protruding into the tree’s xylem (water intake artery).

There, the sensor can monitor plant health indicators such as water flow rate and the relative abundance of important ions, including nitrogen, potassium, and phosphorus. Each device will then wirelessly transmit its sensor data to a central collection hub, such as a nearby research station.

To achieve widespread deployment of their sensors, Liu and Nguyen need to ensure that the devices are inexpensive and durable—and that durability does not only need to apply to the physical components.

“A big weakness of sensors, especially the chemical-related electrical sensors we are using, is that the signal will be degraded over time,” Liu explained. “If you put this sensor in a plant, after one year, the sensor readings will not be reliable at all.”

Liu’s team at the University of Tennessee will be creating a machine learning (ML) model that will continuously recalibrate the sensors from each tree, ensuring that the readings can be trusted years after IoTree devices are installed.

“Theoretically, we can use this machine model to calibrate the sensor quite often—say, every month,” said Liu. “That is a major technical contribution of our approach.”

Replacing Batteries with Trunk Turbines

Another unique aspect of IoTree is the use of self-powered sensors. Many existing plant sensors have hidden costs after deployment due to the need for regular battery replacements.

While Liu’s team works on maintaining long-term sensor reliability, Nguyen’s team at UMass Amherst will be developing the devices’ hardware, including small turbines that will stick out from the bark of the trees.

“We are trying to use wind power to gather energy to support the entire platform,” explained Liu. “That way we can get continuous data from the trees and don’t have to replace batteries.”

To further reduce energy needs, Liu’s team will explore optimizing the sensors for energy-aware computation and allowing them to hibernate when readings are not needed.

“Maybe we don’t need real-time monitoring every second of every day, but just a report once a day or every three days,” Liu said. “In that case, we need to make the sensor hibernate so it’s only activated if people want to read the sensors.”

Impacts that Branch into Other Fields

When the first IoTree sensor prototypes are ready, Liu and Nguyen plan to test them in research groves at UT, UMass Amherst, and a few other universities where they have connections. The team will provide the gathered readings to scientists at each university so the data can be used to inform management decisions immediately.

However, the benefits of IoTree go beyond these small test sites—and even beyond the plant kingdom. Liu and Nguyen are working with open source hardware and software; they plan to distribute the libraries and tutorials made during their grant to boost other scientists working on sustainable monitoring devices.

“This project includes a lot of fundamental research that can be applied to many application domains,” Liu said. “For example, the ML-based calibration model we develop for our chemical sensors can easily extend to other types of sensors, such as those used to measure muscle responses or brain activity.”