AutoWeed develops spraying robot for sugarcane farms

AutoWeed technology has previously targeted weeds in cattle farm pastures and broadacre crops.

Train the weed and crop detection system

In the first year of the two year project AutoWeed from Townsville in Queensland will collect hundreds of thousands of images of sugarcane farmers’ crops, labelled by a human expert, and feed them into deep learning models to train the weed and crop detection system.

The second year will focus on developing and trialling the herbicide delivery component of the project. James Cook University will oversee the development of the smart weed spraying robot.

The project, funded through the partnership between the Great Barrier Reef Foundation and the Australian Government’s Reef Trust, is a collaboration between JCU, AutoWeed, and Sugar Research Australia.

_{Text continues underneath image}

The AutoWeed prototype at work in a crop. – Photo: AutoWeed

AutoWeed spot spraying technology

Jake Wood, AutoWeed co-founder and engineer, says that extending the AutoWeed spot spraying technology to sugarcane requires significant new research and development.

“It will incentivise water quality improvements in reef catchment areas by reducing weed management costs for farmers”, he points out. “While also lowering the concentration of herbicides in runoff to support a healthy reef.”

_{Text continues underneath image}

AutoWeed will sell its system by the metre. – Photo: AutoWeed

The project will run until August 2022. Dr Mostafa Rahimi Azghadi, a senior engineering lecturer at JCU, will lead the project. “We‘re aiming to design, develop, and trial the spot spraying method and fit it to a 24-metre wide, high-rise self-propelled boom to be used on a sugarcane farm”, he explains.

Our system currently sprays with 95% accuracy any weed we train it to spray

Jake says that the system will be able to spray at the root base of the sugarcane in addition to overhead boom spraying for early stage cane.

“And we hope to be able to spray all season. Our system currently sprays with 95% accuracy any weed we train it to spray. This will be the first time a system like ours will be used in sugarcane. The current spot sprayers in the market are only suited to broadacre fallow spraying. And it is an interesting market for us. Sugarcane is a primary industry in this region.”

Green-on-green detection and spray system

AutoWeed has been developing a green-on-green detection and spray system that is applicable to any crop or grazing weed species. The startup has conducted trials across different locations, targeting a dozen significant Australian weed species including Navua sedge, Harrisia cactus, Sowthistle weed, Turnip weed, Chinee apple, Lantana, Parkinsonia, Parthenium, Prickly acacia, Rubber vine, Siam weed and Snake weed.

_{Text continues underneath tweet}

AutoWeed's new green-on-green weed spot-spraying technology is now being deployed to control Navua sedge in pasture on a cattle farm in Malanda QLD. pic.twitter.com/CORvZhRSA3

— AutoWeed (@Auto_Weed) April 16, 2020

Deep learning in real time

The startup uses a combination of in-house electronics and outsourced hardware from trusted manufacturers. “We are using Nvidia processors for example”, Jake says.

“That kind of technology allows us to do some intense deep learning in real time. It is similar to the algorithms that Facebook uses for face recognition. And for the spraying we use custom driver boards coupled with precision fast acting sprayers.”

_{Text continues underneath images}

Drop-in replacement

In a market that knows quite a few competitors – such as Bilberry, Bosch, Blue River and Carbon Bee – AutoWeed offers farmers a system that can be retro-fitted to available agricultural machinery for a reasonable price. “Our system is essentially a drop-in replacement”, Jake says. “It replaces the spraying system they have and can run at any speed that farmers would normally use while spraying.”

_{Text continues underneath video}

The startup aims for a broad market. “A lot of companies focus on just standard cropping”, Jake emphasises. “Our technology also works in-fallow and on pastoral land. We can help dairy farmers for example. They have weed problems as well.”

Farmers are going to make some large savings in herbicide usage

Depending on the infestation AutoWeed can deliver up to a 90% reduction in herbicide usage when compared to traditional blanket spraying, Jake says.

The price of AutoWeed will definitely be a factor in its marketing strategy, Jake expects. “And farmers are going to make some large savings in herbicide usage”, he adds. “It’s also a big positive for farming, when farmers don’t have to use that much herbicide.”

AutoWeed system sold by the metre

AutoWeed will sell its system by the metre. “We can offer a 2 or 3 metre system for pastoral lands but also a 36 metre system for broadacre crops”, Jake explains. “We are developing the system to be modular to make it easily replaceable and serviceable for quick servicing and maintenance. We are developing our units to be robust to reduce any servicing that would be required by our own on-site technicians. Companies that will distribute our products, could also perform the servicing.”

_{Text continues underneath video}

Early adopters

AutoWeed is approaching the end of it’s prototype stage and is working with several early adopters. “We‘re getting closer to full-scale commercialisation”, Jake says. “We are still working on some niceties of the commercial product we want to present when we start rolling out. Within 12 to 18 months we should have our commercial system ready for distribution.”

Built in Townsville, Queensland

The AutoWeed products will be built in Townsville. Initially AutoWeed wants to focus on the market in Queensland. “We have a lot of farming in Queensland, so that market will be the easiest for us starting up”, Jake says. “But we are definitely hoping to go to other farming regions after that. Eventually we could be looking at options overseas.”