Agbot the robotic weed slayer
Agricultural robot Agbot II, designed and built by QUT with support from the Queensland Government, could save Australia’s farm sector $1.3 billion a year by reducing the costs of weeding crops by around 90 per cent.
Professor Tristan Perez, leader of QUT’s agricultural robotics program, said Agbot II’s sensors, software and other electronics enable it to navigate through a field, detect and classify weeds and then kill them either mechanically or chemically. The robot can also be used to apply fertiliser.
“In future versions, the robots could also feed back data on such things as soil and crop health and the state of diseases as they conduct their operations. This would enable better management decisions driven by paddock specific real-time information,”
Darling Downs farmers and the Minister for Agriculture and Fisheries, Leanne Donaldson, had the opportunity to meet the future of farming in Bundaberg in October, when the fully-autonomous Agbot II was demonstrated for the first time.
“The Robotics partnership between the Queensland Government and QUT is a great example of how government and can work together to help our agricultural industries,” the Minister said.
“My Department invested $3 million into this project to help producers use technology to increase efficiencies, profitably and sustainably.
“The Department of Agriculture and Fisheries is committed to investing in innovative technologies to develop Queensland’s agriculture and food industries.”
Professor Perez said Agbot II has demonstrated an outstanding performance in the use of robotic vision and artificial intelligence for the detection and classification of different weed species.
“The cutting edge robotic vision gives Agbot II the ability to spot-spray selected weed species and use mechanical tools to remove other weeds species that are herbicide resistant,” Professor Perez said.
“To date, we have concentrated on the three weeds that are relevant to Queensland: volunteer cotton, sow thistle and wild oats, and the vision system operated with 99% accuracy in the classification of the correct species based on the images collected by the robot cameras.”
He said one of the key problems is that weeds are becoming increasingly immune to chemical control and that was why it was important an agricultural robot could not only detect, but classify the weed species on the spot and decide which actions to take to treat them.
“Agbott ll’s vision system can identify weeds and decide in real time which are better to spray and which are better dealt with, for example, mechanical or thermal methods,” he said.