Yep, I've worked in recycling for 6 years and am well aware that there's a lot of broken parts. My point was that if a PET bottle goes into a single stream system, the recycling industry is pretty good at capturing it, baling it, and selling it to a plastics processor. That processor will clean it, pellet it, and sell it as rPET. 85% of PET bottles that end up in curbside single stream end up getting recycled. I'd like that number higher, but it's where things are at the moment. Throw a bottle in the trash, it's getting landfilled/burned. Throw it in recycling, it's most likely getting recycled.
Based on the single stream comment, I'm going to assume you're describing an experience in North America, and your assumptions are pretty off base for PET bottles.
Figure 1 in the linked paper gives the raw numbers for where PET bottles end up. [1]
I work on a similar type of application (AI connected to a robot arm that sorts recycling). This looks pretty efficient for a few reasons.
- Consistent lighting is crucial for the most efficient AI. Full overhead enclosure makes this way easier
- Gantry style robot is MUCH lighter and easier to repair than both a 6-axis arm or even a delta robot. It's also likely an order of magnitude cheaper than other options
- Gantry robot also makes it pretty modular if they need to modify for different crop widths