The agricultural industry is typically a traditional one that relies on manual labour to get the job done. But researchers at the Australian Centre for Field Robotics (ACFR) at the University of Sydney are working on a robot that could free up time for farmers by killing weeds autonomously.
RIPPA – Robot for Intelligent Perception and Precision Application – has been in development for around three years and includes the VIIPA – Variable Injection Intelligent Precision Applicator, which is mounted to RIPPA and shoots weeds using a micro-dose of liquid.
To navigate around the farm, RIPPA uses GPS. Meanwhile, sensors located underneath it examine the crops and collects data. That data is then passed through various algorithms, such as machine learning and optimisation algorithms, to classify what it sees.
“The output of that training algorithm is an algorithmic model, and that algorithmic model gets thrown back into the robot. The robot then uses that model in real time to determine what is a weed and a crop,” said Salah Sukkarieh, professor of robotics and intelligent systems, School of Aerospace Mechanical and Mechatronic Engineering Director of Research and Innovation at the ACFR.
The omnidirectional robot also runs on solar panels mounted to the top of the robot.
“It can operate for up to 10-12 hours without solar energy, so just on batteries alone. But it’s quite an efficient drive mechanism, so when there’s solar power, it’s going to actually be recharging the batteries at a rate equivalent to how much it loses, so it doesn’t lose energy. We’re aiming for 24/7 operability,” Sukkarieh said.
“We wanted to try something different. We weren’t building a robot that was interested in harvesting or anything like that, so it didn’t require the energy of a harvesting machine.”
Sukkarieh said there are three camps when it comes to farmers using technology on their farm. Sukkarieh said the early adopters will break through and demonstrate how the robot can be used. Next are those that are hesitant but don’t rule it out completely, and finally there are those that have no interest in it whatsoever.
Part of being able to sell the business case to the first two groups included being able to convey how crop intelligence could help farmers. “But it’s very hard to put a value on that,” Sukkarieh said.
But while the agricultural industry can be slow to respond to new technology, Sukkarieh said it is changing with younger growers who are already using digital technology.
“Whether it’s sensor networks or supply chain optimisation, they can see the benefits there already, so … knowing that a robot could have potential is not a big mind shift. How you might use a robot is what they’re focusing on,” he said.
Trials for RIPPA have been conducted on crops in Cowra, NSW, where the robot autonomously drove up and down the rows of crops using satellite-based corrections to within 4 cm precision.
Sukkarieh said the farm had a range of crops, such as spinach and corn, which allowed the team to experiment with and test the robot during different parts of the growing season. This meant they were able to learn how the robot worked in different weather and light conditions.
Sukkarieh said the team learned some valuable lessons during the trials, especially around operational behavior and how much power the robot used.
“The whole thing is computerised and we were measuring everything, so we can actually see where power is being sucked up the most, why it’s being sucked up the most, and under what conditions,” he said.
“We were quite impressed by the whole solar recharging mechanisms, which works quite well now and is practically off-the-shelf almost.”
RIPPA has now reached a commercial prototype stage and the team is looking to start the next phase of development.
“At the moment now we know everything that we need to do in order to build it as a manufacturing run. The next stage of our funding is really looking at doing more trials on farms and closing the loop on farming intelligence,” Sukkarieh said.
This will include working out whether RIPPA can perform multiple tasks at the one time and if it can make decisions based on knowledge it has gathered around the farm.
“We’ve done many things like this before in the past. The challenge that we’re going to face is how do you put into code the transition from information to decision making that a farmer might do,” Sukkarieh said.