Meet Harvey the Harvester: the robot helping sustainable food production

Researchers at the Australian Centre for Robotic Vision at QUT have developed a robot that could help sustainable food production. By 2050, the global population is expected to reach 9 billion and this means a solution is needed to help prevent labour shortages on farms in order to feed this growing population.

Farmers suggested that the cost of harvesting their products outweighs the money they would receive from supermarkets. The new robot could reduce these costs by working autonomously on the farm over a 24-hour period, which would be a huge benefit to the booming capsicum industry in Australia.

“Labour can comprise up to 40% of a farm’s operational costs. The new robots designed and programmed for fruit picking will be used to help with labour shortages and boost sustainable food production for our growing global population,” said Rob Mahony, chief investigator at the Australian Centre for Robotic Vision.

The robot has the vision, understanding and hand-eye coordination to recognise ripe capsicums and gently harvest them. Using deep learning, the robotic arm on wheels, also known as Harvey, uses a camera at the end of its arm to understand the colour and 3D information of a product. Its ‘brain’ is a neural network, which was trained with thousands of images of capsicums and thousands of images of other objects.

Peter Corke, a roboticist and director of the Australian Centre for Robotic Vision, explained that even though green capsicums could be mistaken for green leaves due to their colour, Harvey uses geometry to establish the difference. Once a ripe capsicum has been recognised, the suction cup grips it and uses snippers to cut the stalk.

Under the leadership of Mahony, the centre has said that it plans to build on Harvey’s 90% success rate in initial trials to develop another robot for picking asparagus.

“The Australian Centre for Robotic Vision is excited to be building on the Harvey model to create an advanced asparagus picking robot. There are several challenges in developing and programming robots for farming. The robots must have the ability to see fruit, understand when it is ripe and of course be armed with the hand-eye coordination to pick the fruit without damaging the produce.”

Mahony suggested the new robot would bear many similarities to Harvey in terms of its process and capabilities: “The asparagus picking robot will have a robotic arm for picking fruit and will use advanced 3D computer vision and deep learning to distinguish when and how the produce needs to be picked. We will be developing a neural network trained with thousands of images of asparagus and objects other than asparagus which will act as the robot’s brain helping analyse and understand what it sees.”