hi everyone,
being real, wast sorting bussines isn’t profitable as banking ,thats why we dont see a lot of them near us (what i’m thinking )!!
but the basic idea boils down to number of people running the business less people equals more money
plus builing a recycling center isn’t for small buissiness (do to the profit margin)
how we can solve this ,?
in ordder to grow more recyling bussinesses we need :
less people, more products
ROBOTS !!! hurraaay
the idea is to build a robot capable of sorting plastic from a waste stream and make it affordable for small bussinsses to copy the design .
Any way
there is many waste sorting robots, why we don’t see them. A lot!?
The problem is the cost .Some uses x-ray with fast triggering air nozzles ,some uses arms …
lets go with “arms” one cus i dont like my robot to have xray in his name it sounds frightening (you can think of it like xray need maintence )
the arms i’m thinking is delta parallel robot developed by Reymond Clavel these kind of arms are kown for high speed pick and place which what this company have done .https://www.youtube.com/watch?v=p1rLeNsh06I
don’t worry we still have a place for our non-profit robot design 
how you ask !? this machine costs fortune . SO . How we gona make a machine that cost fortune cost less .
BY making it ourselves hurraaay !!! X2
C:\ TECHNICAL STAFF the robot uses a camera that give a video feed to a computer* running AI program* that detect for example water bottle using image recognitionthen the computer sends a controller command to the arm motors* based on bottle coordination in synch with the conveyor belt speed*,
computer* : needs to be equipped with GPU that is capable of running crysis
AI program* : it will be built using tenser flow (AI lib open to the public by google ) Building a CNN (conventional neural network) model will be nessecery to differentiate between all kinds bottles so that they can be sorted.
Arm motors*: most delta robot uses servo motor duo to high speed/torque capabilities.
Speed* : conveyor belt speed sensor is essentiel to keep the robot accurate ,to be able to calculate the delay between the video detection and the picking action.
love to hear feed back, question in any way shape or form . here is my email ouahhidi@gmail.com or in comment
it would be awesome if i got accepted to contribute this idea in the devolepment of Precious Plastic Version 4
i need help any way i don’t think i can pull-it on my own
By the way if we find some geek who can convert our trained cnn model to A FPGA ship , it will be much cost effective than using a computer with high end gpu
a cnn model can be trained on any object with specifice shape and color (card board , milk jugs ,… )
you can imagine multiple robot on the same conveyor belt .
0
