Niagara College is at the forefront of agricultural robotics technology, with its complement of aerial drones and enterprise-level software systems. Now a 400-lb remotely operated rover, named RoamIO Jumbo, is undergoing field testing to refine high-precision engineering onboard that will help accurately predict vineyard yield estimates for grape growers.
These field trials are taking place at the Niagara-on-the-Lake campus vineyards by members of the Agriculture & Environmental Technologies Innovation Centre (AETIC) research team, after months of in-lab engineering. The technical advancement project, which also includes optimizing autonomous navigation capabilities, is in collaboration with industry partner Korechi Innovations, the creators of the RoamIO robot.
RoamIO is outfitted with various sensors, including traditional image capturing cameras, LiDARs, SONARS, temperature probes, soil moisture probes and more. All these sensors will help grape farmers increase profitability or even help save their crops by better understanding the total expected yield and the variability across the vineyard.
Thanks to a federal grant from the Natural Sciences and Engineering Research Council of Canada (NSERC), the College was able to purchase RoamIO last year, and is now working with Korechi to optimize the robot’s autonomous navigation and spatial awareness capabilities and enable it with computer vision to obtain accurate and variable yield estimates.
The first phase of the agricultural robot project involved the AETIC team working with Korechi’s original onboard software to refine the auto-navigation programming using waypoint GPS technology to improve the rover speed, rather than having it stop to recalibrate for position for a fraction of a second every 10 metres.
“The College team implemented 2D LiDAR because it has a higher degree of precision for obstacle detection and collision avoidance,” says Sougata Pahari, Founder & CEO of Korechi. “Niagara College has been instrumental in solving these problems and others are in progress.”
For the navigation system, the research team works collaboratively with Korechi’s engineers – including two full-time software coders – who also provide rigorous testing on the work carried out by NC for validation.
Pahari is already bringing this auto-navigation capability to the market, giving demonstrations using other land rovers his team has built. His target market is 20- to 200-acre vineyards, with a secondary market of golf courses, as the enhanced platform allows better performance in open fields. He’s also working on a software app to be used on a tablet.
“We’re not talking about 100 percent autonomy at the moment and that’s probably for the best, because farmers want to have a say in how their farms are operated,” notes Pahari.
The second and current phase of the project brings in artificial intelligence (AI) to capture specialized views of a vineyard, and different types and colours of grapes, to get an accurate and variable yield estimate.
Typically, a grape harvest yield prediction involves workers taking samples by hand from areas of a vineyard and then making an educated guess for the entire field.
“The problem is accuracy: while humans can only get within 25 percent error, machine vision, however, can get within 10 percent error,” explains Andrew Nickel, Research Associate and a NC Electronics Engineering Technology program graduate who is overseeing the multi-phase project.
There is also the challenge of yield varying considerably from year to year and also geographically within the field due to soil conditions, pests, and climate. And Niagara is more susceptible to this than most grape growing regions, says Nickel, who is also a former grape vineyard owner.
“In Niagara we have always had variable growing seasons, whereas winemaking in Chile or Australia is like Groundhog Day; every season is the same.”
Instead, the innovative technology being created within the Research & Innovation division involves designing the vision system using commercial hardware and camera systems. It’s a challenging endeavour to program such precise machine learning algorithms for grape identification, colour and counting given the array of variables, such as differing grape size, cluster shapes and lighting conditions.
Once complete, the field data will give farmers a window into the total expected yield and variability across the vineyard, allowing for more accurate sales and revenue prediction. The farmers can also use the data to better manage their vineyards by using measures such as variable rate fertilization and variable rate fungicide application to improve their yield.
“It’s been a tremendous partnership and Niagara College is such a great institution that it lends more legitimacy whenever we go talk to potential customers,” adds Pahari. “Everyone recognizes Niagara College as one of the leaders of knowledge in the vineyard space in Canada.”
The Agriculture & Environmental Technologies Innovation Centre (AETIC) research team is comprised of computer programming, electronics, robotics, and geospatial information systems (FIS) students, and recent graduates. The team is led by Mike Duncan, PhD, who is in his eighth year as the Natural Sciences and Engineering Research Council’s Industrial Research Chair for Colleges (NSERC-IRCC) in Precision Agriculture & Environmental Technologies at the College.