Given that phones comes with FaceID for cheap 3D scanning, and the level of object recognition that comes with deep neural networks these days, I'm curious what major obstacles remain?
Surely pads on robot fingers can ensure soft and distributed enough pressure? Is it something to do with holding the stem with one robot grip while pulling the strawberry? Is it pulling apart parts of the plant systematically to find the strawberries in the first place? Or the fact that the plant isn't rigid like a tree, but that the location of the strawberries depends on how it's being held or pulled apart in the first place?
Also curious what the break-even point is, where even if it collects less strawberries than people do, it still winds up being cheaper.
Kind of bummed the article doesn't go any details at all except for avoiding the half-eaten ones.
I picked strawberries as a kid (we'd go to the local orchard) and it was so much fun, but I can't really remember the mechanics of it.
Detecting strawberries is easy, but there are several factors once you've done that.
You need to determine the location, the orientation to determine the best cutting angle (and then finding a motion solution for the robot). You need to determine if the fruit is occluded and if you can actually reach it. A human will instinctively move foliage out of the way. A robot has to have some way of doing that without destroying the plant.
Different companies have different approaches. I really like the agrobot method which cups the strawberry and then pulls it upwards, severing the stem. Grasping is not viable right now, in my opinion, it's too easy to damage the crop. There's lots of research going on (force-torque sensors, adaptive/smart grasping), but it's still in universities.
The issue is that there are actually many implicit constraints on the motion plans by which robots would pick strawberries, and when you make any one of them explicit, your whole robots solution might have to be redesigned from the ground up.
Surely pads on robot fingers can ensure soft and distributed enough pressure? Is it something to do with holding the stem with one robot grip while pulling the strawberry? Is it pulling apart parts of the plant systematically to find the strawberries in the first place? Or the fact that the plant isn't rigid like a tree, but that the location of the strawberries depends on how it's being held or pulled apart in the first place?
Also curious what the break-even point is, where even if it collects less strawberries than people do, it still winds up being cheaper.
Kind of bummed the article doesn't go any details at all except for avoiding the half-eaten ones.
I picked strawberries as a kid (we'd go to the local orchard) and it was so much fun, but I can't really remember the mechanics of it.