AI is reshaping agriculture. A new robot roguer designed for Potato Virus Y detection could revolutionize how farmers manage potato diseases.
Disease rogueing is a notoriously labour-intensive effort. What if an AI-enabled robot could do the work instead, better and faster than any set of human eyes and hands? That’s exactly what Charanpreet Singh, a master’s student at the University of Prince Edward Island, wondered when he set out to build a robotic roguer capable of detecting Potato Virus Y (PVY). Now in field trials, his robot may be the start of a whole new kind of field management.
“Before starting my masters, I worked with an AgTech startup developing a robot for weeding,” says Singh. “In 2023, when I began my masters at UPEI, my supervisor, Dr. Aitazaz Farooque, introduced me to the challenges growers face in managing PVY. Detecting PVY symptoms on potato leaves requires high-resolution images under controlled lighting conditions. We explored various solutions, including capturing images with drones, handheld cameras, and cameras mounted on farm equipment. However, the images from drones and mounted cameras were not high quality enough to detect the subtle symptoms of PVY. After several brainstorming sessions, we concluded that developing a specialized robot was the best approach.”
AI Rolling into Farmers’ Fields
Artificial intelligence (AI) is sweeping across the world and changing the way many people work across countless industries, agriculture included. Already, AI is being used in several innovative ways to improve efficiency, crop quality and sustainability, from nutrient deficiency identification, to yield prediction, to precision input application and more.
It’s also being used on the pest and disease front, where AI systems working together with drones and ground sensors can scan potato fields to detect early symptoms of infestation or outbreak.
Singh’s robotic roguer was tested in grower fields (both seed and commercial) and in plots at multiple research farms on P.E.I. and in New Brunswick in both 2023 and 2024. Built with the support of the PEI Potato Board and other partners, the robot has proven impressively successful: in fact, it is up to 90 per cent effective in detecting the unique presenting mosaic of PVY symptoms in plants across a range of potato varieties.
Now in its second year, this project is a collaborative effort involving multiple provinces and organizations, including the PEI Potato Board, PEI Department of Agriculture, Potatoes New Brunswick, New Brunswick Department of Agriculture, Aquaculture and Fisheries, and UPEI. The robotic system, which was developed from work done by another UPEI student who successfully employed machine vision to detect early blight symptoms and lambs quarters, is armed with five cameras and satellite internet. The current iteration of the robot can scan five rows at a time, photographing plants then, due to limitations with onboard power and computing, sending the geotagged files off to be analyzed remotely by the AI model.
“The robot uses a database of images to identify PVY symptoms and tag them as infected,” project collaborator Ryan Barrett, research and agronomy specialist at the PEI Potato Board, notes.
Once the model flags potential PVY infections it generates what Aitazaz Farooque, associate dean of UPEI’s School of Climate Change and Adaptation calls an “infection map”, with the infected plants geotagged so that they can be verified and then pulled. Farooque says the goal is “to add a paint spray mechanism so that it’s easier to identify where those infected plants are.”
Big Steps Forward
PVY has plagued the potato industry and others for decades, devastating crop yields and costing farmers dearly.
Making the problem of PVY even harder to manage is the growing challenge of hiring seasonal roguers willing to walk fields in search of infected plants.
According to Barrett, newer strains of PVY are becoming increasingly hard to identify by eye and, with fewer experienced roguers available and even fewer able to successfully do the job, finding and removing infected plants is becoming more and more troublesome for growers.
“When the chance arose to apply this technology to potatoes, we recognized an opportunity to tackle a significant challenge in the industry,” Barrett says.
This technology has the potential to revolutionize the potato industry, saving both time and money. Barrett emphasizes the benefits of the system: “The robot not only reduces reliance on human labour but also performs the task with greater precision.”
For example, in one field test the robot was able to identify an infected plant nearly hidden among healthy ones — a situation where a human roguer might easily have missed the infection.
“The robot geotagged the location, and when we revisited the spot, it was clear the robot had correctly identified a PVY infection that might have gone unnoticed,” Barrett explains.
In its first year, the AI model based its predictions on a bank of less than 100,000 images. Now, nearly 1,000,000 images are banked, dramatically building the system’s accuracy.
“The more data the system processes, the better it becomes at identifying PVY symptoms,” Barrett says. “This continuous learning process ensures that the technology will only become more valuable for farmers.”
“PVY symptoms vary depending on the potato variety,” adds Singh. “Currently, our algorithm works on a few varieties, and as we gather more images from new varieties, our algorithm can learn to detect symptoms in those as well.”
Commercialization Ahead?
As the project advances, the research and development team is already considering future applications and enhancements, including adding solar panels that would extend the robot’s operational time in the field without recharging batteries.
The team would also like to make it operate fully autonomously based on an uploaded field map. According to Farooque, farmers might in the near future be able to simply use a touch screen interface to pick a potato, be it Yukon Gold, Russet Burbank, etc., then send out the robot to fully independently rove that specific field in search of infected plants.
Ultimately, Barrett says, “Our goal is to develop a platform that can be commercialized, either as a complete system or as a standalone AI model that could be integrated into existing farm equipment.”
There are still several steps necessary before commercialization, Singh says.
“Before commercializing, we need to conduct further testing over the next few years to evaluate the models’ performance across different potato varieties and to increase the runtime so the robot can cover more acres,” Singh adds. “From the growers’ perspective, they need to rogue hundreds of acres within a two-to-three-week window.”
He also would like to shift the mapping process to real-time detection and identification.
“At present, we process images post-capture to detect PVY symptoms, which means the infestation map is generated after the robot scans the entire field. In the future, we aim to achieve real-time detection,” he says.
In the longer term, potential applications for the technology are vast. As the project continues to evolve, it offers a glimpse into a future where AI and robotics play integral roles in agriculture, helping farmers overcome challenges that have long ranged from frustrating, to costly, to nearly insurmountable. AI could set a new standard in precision agriculture, with implications that reach well beyond potatoes. The ability to detect subtle symptoms and trends that might elude even the most trained human eye could lead to breakthroughs in managing various plant diseases, ultimately contributing to more sustainable and productive farming practices worldwide. Using AI to monitor plant health, farmers could detect diseases earlier, apply treatments more precisely, and ultimately increase their yields while reducing chemical use.
“The platform we developed can detect any disease that shows visible symptoms,” says Singh. “We can train our AI models to detect new diseases with a dataset of images showing the symptoms. Currently, we are working with growers from P.E.I. and New Brunswick, focusing primarily on potatoes since they are a major crop in these regions. In the future, we would love to collaborate with growers of other crops as well.”
Looking ahead, the team aims to refine the technology further and expand its capabilities. They are focused on enhancing the AI model’s accuracy, improving the robot’s ability to navigate various field conditions, and making the system more user-friendly for farmers.
Barrett is optimistic. “We’re still in the trial phase, but the results so far are promising,” he says. “With continued development, this AI-driven approach could become an essential tool in the fight against PVY, helping secure the future of potato farming in Prince Edward Island and beyond.”
For Singh, who grew up on a farm and feels a deep connection to agriculture, “it feels incredible to see [the robot] working in real fields, especially when we receive positive feedback from growers about its usefulness. From day one, our goal was to design a robust and durable robot that could withstand challenging field conditions, and seeing it perform well in those conditions is very rewarding.”
