This is the next phase of a voice-assisted farm management system showcased by researchers at this year’s Mystery Creek Fieldays.
While its title of “Smart Assistant” is simple enough, the interactive artificial intelligence system incorporates gigabytes of data collected from the university’s No.4 dairy farm, and accessible through Amazon’s Alexa voice device.
“What we found in the course of developing the Alexa-driven system was farmers telling us there was no shortage of data out there for them,” Johan Potgieter, professor of robotics, says.
“Their issue was accessing what they wanted in a simpler way. Even having apps now, there are simply too many of them and finding the right app, the right dashboard of information is now too much to access. They wanted their information to be more digestible.”
Rather, they were seeking information from that data that highlighted the abnormal, the departures from the expected.
“But given the level of volatility farms now experience, it can be hard for us as humans to catch those anomalies, whereas Artificial Intelligence (AI) engines are very good at this.
“Deep learning systems absorb all the data, they can ‘remember’ previous anomalies, and find links between them, making sense of that data better than we are capable of doing.”
The Massey farm assistant proved capable through its Alexa voice interface of answering questions as broad as ‘how milking went’ or “what is the cover in paddock xy?’.
Potgieter said the Massey system was an effort to take AI and machine learning out of the realm of pure robotics and into the agri-space.
“Initially machine learning started through surveillance systems analysing videos of crowds for security.”
But major advances in open platform, cloud-based AI software have meant technology can be applied in more intuitive interactive uses.
The smarts behind the Massey system is TensorFlow, an open-source AI software library developed by Google that provides super computer capacity to researchers wanting to advance AI-driven systems without having to invest in the computer power themselves.
“We basically feed all the information and images we have into this, with the parameters and weightings we want.”
New Zealand traditionally has a poor reputation for tech transfer from concept to commercial pick up.
“So our hope is that by using these open source platforms that are accessible for researchers it will hasten that pick up through to the likes of farmers.”
Initial responses from farmers at both Mystery Creek and Central Districts Fieldays have been overwhelmingly positive.
“They like it that all those apps they have to access are all sitting behind Alexa, and it’s Alexa that has to use them to get the information they are seeking – you don’t have to open them yourself. There is no password or dashboard, you simply receive the answer you want.”
Grumbling opposition occasionally comes from farmers at the baby boomer age, but younger farmers are fascinated and keen to try the system out.
The researchers anticipate commercialisation of the system is close, but what particularly excites Potgieter is the next phase of the project.
“That is where the system starts to make recommendations to you. That could be to turn on an irrigator, shift the stock or start harvesting.”
Using an array of sensors linked via the internet of things that include monitors for weather, soil moisture, water flows, pasture growth and stock demand, he intends to move it to that level.
“Really this is as close as we can find a volunteer farm to set it up on as a trial adopter.”
