Employing AI to Sort Plastic Waste by Manufacturer

Faculty Lead: Bradley D. Olsen, Alexander and I. Michael Kasser (1960) Professor of Chemical Engineering at MIT

The vast majority of plastic products and packaging sent to materials recovery facilities are multi-material or contain additives such as dyes, making it challenging to recover high purity bales of a single type of plastic. Even as we move towards mono-material, the mechanical and rheological properties of the specific grade of plastic used for a product are unknown. However, if plastic producers use recycled content from their own post-consumer products and packaging, it can greatly mitigate this issue, making mechanical recycling even more feasible. As plastics are already highly efficiently “barcoded” based on branding used to promote consumer recognition, this project will use these branding marks to effectively sort plastics by manufacturer type instead of plastic type using AI image recognition. This will enable manufacturers to either buy back or take back in exchange for a tax rebate their own products and use them most efficiently in recycling, as the grade and additives of the original product will be a match to the grade and additives of the new product. This workflow is well-suited for Extended Producer Responsibility (EPR) which places the onus of packaging recyclability on the producer or manufacturer and presents a potentially more effective and palatable alternative to the general plastic production taxes or bans being implemented in jurisdictions today. Because this technology could also be readily implemented with modest modifications to current system operations and without system-wide collective agreement on change, it presents an extremely promising approach for near-term action to improve circularity.