Ever walked in an orchard and, feeling a bit like Adam, picked up a nice apple? What could be easier than that. You see the tree, all the nice apples hanging around, you just need to choose the one looking nicer, and probably tastier , stretch your arm and you pick it up.
Actually, that is one of the most complex activities you can imagine!
Unperceived by you, your eye, brain and body muscles enter into a tight coordination process involving sophisticated decisions that leverage on plenty of past experience.
You first have to spot the apple tree, then look at the branches and spot the apples distinguishing them from leaves and may be from a plastic ball that some kids throw on the tree. Then you have to look for the ones that are ripe, a nice reddish hue, or may be a yellow one if that is that kind of apple…, and make up your mind which one to pick. At this point it becomes a matter of gauging the position of the apple with respect to your position, getting closer to the point that your extended arm bring the apple in the vicinity of your hand. Then you have to move your hand in a 3D space, avoiding branches and picking up the apple with just the right pressure, twisting it to detach the stem from the tree.
There are visual recognition and spatial recognition capabilities involved and they are way tougher that you might perceive. It comes natural to us, but that is after months, years of practicing and learning these capabilities in our infancy.
Now, try to teach this to a robot!
Several companies are at work to do exactly this: develop a robot that can pick up apples. Anf pick them up choosing the ones that a ripe without damaging the tree and the other apples that are ripening,.
One company, Abundant Robotics, seems to have managed this, although they are making their life easier by customising the robot to operate in New Zealand orchards. These orchards have been “engineered” by farmers to look like vineyard, with the apple trees aligned in rows, well-spaced from one another, and with the trees looking more like two dimensional than three dimensional trees.
Still, it takes quite a bit of technology to spot the apples, select the ones ready to harvest, and pick them up with a pneumatic tube that using the right sucking pressure detaches the apple from the branch and puts it into a basket. All of this without damaging the tree nor disturbing the other apples (watch the clip).
What amazes me is the hidden complexity in actions that we continuously perform without paying a second thought. As soon as you try to have these actions replicated by a machine you realise their complexity. Abundant Robotics is using sophisticated sensors, including LIDAR, and artificial intelligent based computer vision. There is also quite a bit of sophisticate and precision mechanics involved to manage the pick up pneumatic tube.
Agriculture is moving towards a new revolution, after the one brought by mechanisation of work. The new wave is involving autonomous systems, in several cases coordinating activities among them, some operating on the ground, others flying over the field, monitoring and spraying fertiliser and pesticide.