Robot harvests lettuce, and that’s impressive
Machine learning may boost automation of agriculture. Nick Carne reports.
A British robot has harvested a lettuce. That might, at first glance, not appear terribly exciting, but it could represent an important advance in the automation of agriculture.
While crops such as potatoes and wheat have been harvested mechanically at scale for decades, others defy progress and still need to be picked by hand.
The iceberg lettuce – the most common type grown in the UK – is one such beast. It is easily damaged and grows relatively flat on the ground, presenting a challenge for robotic harvesters.
But maybe not for the “Vegebot”, which uses machine learning to identify and harvest its lettuce.
Developed by engineers at the University of Cambridge, it initially was trained to hunt and gather in a lab setting, but has now also been successfully tested in a variety of field conditions.
It is as yet nowhere near as fast or efficient as a human picker, but its potential excites Simon Birrell and his colleagues.
"Every field is different, every lettuce is different," he says. "But if we can make a robotic harvester work with iceberg lettuce, we could also make it work with many other crops."
The Vegebot has two main components: a computer vision system and a cutting system.
An overhead camera first takes an image of the lettuce field, identifies all the lettuces in the image and determines which should be harvested. A lettuce may be rejected because it isn’t mature enough or has a disease that could spread to others in the harvest.
A second camera near the cutting blade helps ensure a smooth cut. The pressure in the robot's gripping arm can be adjusted so it holds the lettuce firmly, but without crushing it.
The key part of the work was developing and training a machine learning algorithm to recognise images of healthy lettuces in the lab, then real lettuces in the field, in a variety of weather conditions.
"We wanted to develop approaches that weren't necessarily specific to iceberg lettuce, so that they can be used for other types of above-ground crops," says research leader Fumiya Iida.
The researchers acknowledge that Vegebot has to get an awful lot quicker to be of practical use, but say it is not just a direct question of robot versus human.
In future, robotic harvesters could address problems with labour shortages and even help reduce food waste, they suggest.
At the moment, each field is typically harvested once, and unripe vegetables or fruits are discarded. However, a robotic harvester could perform multiple passes on the same field, returning at a later date to harvest those that were left during previous passes.
The study’s findings are published in The Journal of Field Robotics.