Glacier FarmMedia – The weed control equipment trundling through the field in southern Ontario checked a lot of sci-fi boxes.
They were robots, used artificial intelligence and killed weeds with laser beams.
Field-applied plant cell exploding technology was one of the cutting edge agricultural ideas presented at the Bradford AgRobotics demonstration day earlier this summer.
Read Also
Agco autonomous farm equipment rig ready for limited release
Agco’s PTx Trimble has made a few changes to its Outrun autonomous grain cart and tillage system in the past year, in which the company has “taken that system and converted it to look at it as an autonomy platform.”
Tech entrepreneur David Tao, chief executive officer and co-founder of Ontario-based BH Frontier Solutions, presented a compact BHF ElectricWeeder, while Dutch ag-tech firm Pixelfarming demonstrated a modified laser model.
Haggerty AgRobotics president Chuck Baresich said the Pixelfarming Laser One-i prototype, adapted for the Orio NAiO’s frame, arrived in March. NAiO is an autonomous field robot in its third year at the Ontario Crops Research Centre in Bradford.
It features six laser units that use high voltage to energize a carbon dioxide tube, creating a focused beam that boils the liquid inside plant cells, effectively destroying the plant.
“The lasers will shoot forward and back,” Baresich said. “It will determine the optimum angle as you’re moving forward to hit the various weeds.”
Rydan Chalmers, a Haggerty AgRobotics Systems Engineering intern, used 35 images of multicoloured confetti to demonstrate how easy it is to program the interface’s crop module to target specific plants or, in this case, confetti colours for eradication.
Baresich clarified that plant images at various growth stages and under different lighting conditions are necessary to develop a comprehensive crop model, noting that the rapid pace of technological advancement and the availability of open-source tools are accelerating the process.
“You can load five (models) at a time,” Chalmers said.
“We have a few examples we’ve been testing in the field. And you want to make sure you have the right laser selected, (and) how much energy is going to each laser.”
For the demonstration, Chalmers set the lasers to 80 per cent, which, while producing a satisfying flame-out, was excessively high because most weeds and grasses are effectively terminated at 25 to 40 per cent power.
Baresich warned that there is a low chance of starting a small fire if a field is dry or of damaging a neighbouring crop if the heat isn’t properly adjusted.
The LaserOne-i’s speed and effectiveness are optimized for dime-sized plants or early emergence because larger plants slow the machine significantly.
“When you have a very, very tiny weed, it takes only milliseconds of heat to cook that,” Baresich said.
“That’s how you get the speed of the machine up.”
Field trials comparing 90,000 weeds to a million weeds per acre showed a significant slowing of the machine due to the number of shots per second per acre, a common issue when dealing with grass because it’s prolific, Baresich said.
Depending on the heat level and dwell time required to eliminate a plant, the three lithium phosphate batteries provide a four-hour run time, with the ability to add additional batteries.
BHF ElectricWeeder is an autonomous robot that selectively kills weeds with high-voltage electricity. It is guided by artificial intelligence using an integrated multi-camera computer algorithm that differentiates between weeds and crops.
“The algorithms are trained to recognize the crop; anything different from that would be considered a weed,” Tao said, adding that the crop model trains with 100 images, while shape and morphology changes require additional training.
The robot collects data automatically, improving the AI’s accuracy and increasing its speed without the need for farmer oversight outside of the initial RTK to plot out the rows.
“Once it detects the weeds, it will guide the robot arms underneath and basically target the weed and zap it,” said Tao, adding it’s part of an onion crop trial at the research centre.
During the demo, the robotic arms moved swiftly and methodically through the crop, unleashing an electric current that often produced a visible puff of vapour as the weed’s cells burst.
The key difference between the ElectricWeeder and other robotic options is its ability to dispatch weeds at any size and growth stage.
“It has two modes of operation tailored towards both low weed density fields and high weed density fields,” Tao said.
“It will maintain the same efficiency at killing weeds.”
The 1.2-tonne machine can weed up to 10 acres an hour and operate in most weather conditions 24/7, but like traditional iron, mud can be problematic.
“A bit of rain, a bit of drizzle, a bit of moisture is fine, but not too much,” he said.
“Not like a muddy kind of pouring rain where you have ponds in the soil.”
The ElectricWeeder comes in autonomous, tractor-pulled and tractor-liftable models, with coverage from five to 20 feet and swath widths of 60, 120 and 240 inches, plus an adjustable row spacing of 60 to 88 inches. The sub-millimetre precision and 99 per cent weed kill rate make it more efficient than a 100 person hand crew.
Tao said the US$150,000 commercial versions will ship near the end of the year.
Robot weed control will never be a one-time pass, said Baresich.
“Part of the economics of the machine is how many times do I have to go through that same field, and can I get it done?” he said.
“In comparison to hand-weeding crews, where the costs are $500 to $1,000 an acre, these are very economical to run.”
He said chemical weed treatment remains cheaper; but in cases of herbicide resistance and hand pulling, the return on investment of laser or electric weeders pencils out.
“I think the reality is that even though the ROI might not work, we’re going to be forced into the robotic technology just because of weed resistance,” said Baresich.
