Recycling centers use AI to identify and sort

AI-driven systems can help with recycling, that requires rapid identification of objects with diverse shapes, sizes and orientations on conveyor belts. (

DEAR EARTH TALK: How are AI and robotics combining to revolutionize recycling?

-- B.C.,

New York

ET: A survey conducted by the Carton Council of North America in 2018 showed that 94% of Americans support recycling. That same year, the Environmental Protection Agency reported that the recycling rate was only 32.1%. Why is this the case?

Local governments are responsible for creating recycling programs. Cities that have invested in recycling infrastructure, education and incentive programs, such as San Francisco and Los Angeles, claim recycling rates of more than 70%. Cities with smaller budgets and staff have eliminated curbside recycling altogether. (Chesapeake, Va., and Pembroke Pines, Fla., are two examples.)

Single-stream recycling, where various recyclables go in a single container, has increased household participation significantly. But it has also contributed to a 25% contamination rate. Contamination occurs when non-recyclable items are mixed with recyclables, making it challenging or impossible to sort and safely process. Common contaminants include non-recyclable plastics (bubble wrap, trash bags, cling wrap, etc.) and food residue.

Contamination is more than an inconvenience. In 2016, China received more than 16 million tons of plastic, paper and metals from the United States, 30% of which was contaminated and later dumped in the Chinese countryside and waterways. In 2017, China passed the National Sword Policy, banning materials that the U.S. had previously sent in for recycling. As a result, U.S. facilities have had to make substantial improvements in the quality of their recyclables.

How does AI play a role in improving recycling? The 1990s saw the introduction of optical sensing and computational intelligence to distinguish between types of plastic and paper. These systems typically achieved 80% to 95% purity, while human workers remove contaminants manually. Enter artificial intelligence. Recycling requires rapid identification of objects with diverse shapes, sizes and orientations on conveyor belts. AI-driven systems demonstrate near-100% accuracy by relying on image analysis of attributes, including color, opacity and form. A vast dataset of recyclable material images are regularly updated to improve reliability.

One company, AMP Robotics, has pioneered in the AI-recycling industry since 2014. Equipped with a powerful network, their 1,800-pound "pick-and-place" robots are twice as efficient as human employees. Now recycling facilities equipped with artificial intelligence robots are able to sort greater quantities of trash while reducing operating costs.

Perhaps we can even stop contamination at the point of disposal, right at home. CleanRobotics has created a receptacle named TrashBot that uses imaging, AI algorithms, and robotics to detect and sort waste as it is being thrown away. This makes the sorting process easier down the line.

EarthTalk is produced by Roddy Scheer & Doug Moss for the 501(c)3 nonprofit EarthTalk.