I should confess that when I read this news, a robot harvesting raspberry, what impressed me most was the statement that a human “harvester” could pick up 15,000 berries in an 8 hours shift. I would have imagined, based on personal raspberry harvesting experience, so much less (may be that’s because I equally divide my time between picking them up and eating them…).
Fieldwork Robotics, a spin off of the Plymouth University, is focussing of robots for use in agriculture and they decided to take the hardest road, probably because there are already a few successful companies offering robots for seeding, tilling, harvesting of maze, wheat … in a way all the “low hanging fruits” business is already taken.
Picking up raspberry is difficult, both because the shape of the plant (a bush actually) and the position of the berries changes a lot between plants and in the same plant, with leaves covering part of the berries, and because the berries are soft but they are also sticking to the stem so that the harvesting arm needs to apply the right force, nothing less and nothing more.
They have addressed the problem of detecting the position of the berries using, you guessed already, artificial intelligence algorithms to process the images captured by a camera on the robotic arm, and sensors coupled with actuators through a smart algorithm to apply the right pressure.
The robotic arm moves around on wheels and it is able to pick up over 25,000 berries a day in an 8 hours shift (10,000 more than an expert human harvester) and, of course, the robot is not going to complain if its working hours get extended! This is what it is being claimed. Looking at the clip (see below) it would not seem to me that quick at all (in the time it took it to pick up the berry I would have eaten quite a few of them!) but this is a prototype used to demonstrate the capability of the computer vision algorithm to spot a ripe raspberry among leaves and unripe berries.
At the end of May Fieldwork Robotics has completed the trials and is now moving the next step towards industrialisation, expecting to deliver the commercial version early 2020, before the harvest period. At the same time it will be working on creating other robots to harvest different types of fruits (in most cases this will involve a change in the software part, that is the retraining of the computer vision algorithms to recognise different types of fruits, the pinchers and the movement control will need to be adapted as well).
If you like to see a current industrial harvester that is harvesting many thousands berries take a look at the second clip! Notice that this “mechanical” harvester requires that the “berry” field has been laid out i a very precise way, which is not the case of the Fieldwork Robot that can operate in any environment.
And her is the mechanical harvester….