Autonomous Citrus Harvester

24/03/21 01:15 PM Comment(s) By FIAL

UBOT | Celebrating Australian Food and Agribusiness Innovations 2020

Challenge

When Patrick Edwards decided to move from IT into robotics, he was looking for a problem to solve. Seeing the challenges faced by Riverina citrus growers β€” specifically the difficulties in acquiring and cultivating a dependable workforce to harvest produce β€” he knew he could help.

 

It’s an undisputable fact that no-one likes to pick oranges. Compounding that issue, growers compete with one another for harvest labour, drawn from a small potential pool of itinerant workers, hindering their ability to efficiently bring product to market.

 

Solution

Patrick knew solving the problem meant bettering human labour results, so he set about developing a robotic harvester that offered an alternative to manual picking, inclusive of object detection capability.

 

Early iterations utilised SVM artificial intelligence, which offered good detection results but required a lot of training to handle variations in conditions like light levels and shadows. The project stalled for a time, after which UBOT implemented r-cnn object detection, which was an order of magnitude greater and produced far better results. A team of Master of Data Science students at James Cook University have been developing a smaller AI object detection model with promising results β€” a model using electronics and drawing less power will allow the robot to run for longer.

 

Keeping the build cost down has been a priority for UBOT, knowing that the market for million-dollar robots with a 20-year ROI would be non-existent. He used off-the-shelf motors and electronics to facilitate that endeavour, with the latest version offering a return in as little as one year.

 

Outcome

Commercialisation is close, with a new beta version currently being built. The ultimate goal is to run robots 24 hours a day, seven days a week, giving farmers a low-cost alternative to an unpredictable labour force. While the local market is relatively small, growers have suggested they could easily double their acreage if not for labour limitations. Seems like UBOT is just what the citrus industry needs.

FIAL