AI and IoT Target the Flow of Plastics in Waterways

Published: July 15, 2024
  • IoT devices, using LoRaWAN and Wi-Fi technology, capture and transmit sensor data about contaminants in recycling bins as well as the amount of biowaste collecting in bins.
  • Saint Louis University initiative tracks recycling and ensures that plastics that should be recycled don’t end up in landfills.

Saint Louis University (SLU) researchers are expanding an IoT-based pilot, in the alleyways of St. Louis, that identifies and tracks the volume and type of household recycling and waste that are thrown into bins. The goal is to better understand when recycling is being contaminated with waste, or when plastics and other products, that should be recycled, end up in waste bins, and then landfills.

The solution uses solar-powered devices inside bins, that send sensor readings via an IoT system from waste bins, including optical data from inside recycling bins.  The lab where the solution was built is known as artificial intelligence for coupled human environment system analysis for sustainability (AI-CHESS).

That AI functionality can help identify objects in bins to determine what has been thrown away and when, explained Orhun Aydin, project lead, and SLU’s assistant professor in the Earth and Atmospheric Sciences Department.

The project is funded in part by the National Science Foundation (NSF), while the research team is working together with Earth Day 365, a nonprofit aimed at environmental sustainability in St. Louis. Earth Day 365 provides the community outreach related to data from the system, and works with local residents. The goal is to achieve optimum recycling rates without overburdening people, said Aydin.

Turning off the Flow if Plastics Into Rivers, Seas

Aydin has a background in a geo-statistics discipline known as geospatial artificial intelligence (GeoAI) and the IoT project stems from his early work in measuring the flow of plastics in a place they should not be in earth’s waterways.

He has worked on time-modeling of plastic flows from rivers into oceans as well as modeling plastic islands that float in the open sea. There are currently about 2.25 trillion pieces of plastic debris in the ocean and every year about 8 million tons of plastic waste escapes into our waterways.

“Looking at the scale of this problem I wondered what can we do to turn off the tap of plastics at the rivers or before [plastics] get that far,” said Aydin.

Satellite data is useful, but still too coarse to detect plastics in the oceans and the rivers. Additionally, camera systems along the river do help count plastics, but gaining a wide scale view at the riverside would be too costly.

 Looking in the Bins

Long before discarded plastic reaches the river, said Aydin, there is a story of mismanagement in which plastics that were supposed to be recycled somehow wasn’t. The smart bin solution aims to take the view further upstream to the place where finished plastics often end up: the bins around people’s homes.

Looking for ways to close off the plastics tap will require better management of recycling, Aydin and his team posited. But understanding how the recycling breakdown happens is a challenge.

“We have a huge data gap once the waste leaves the consumer, once it leaves our hand and goes into a bin somewhere we don’t really know what happens to it,” Aydin said.

A Pilot to Track Recycling with AI

Residents of St. Louis use alley bins—communal neighborhood receptacles in which multiple households dispose of their recycling and waste. When things go right, one bin fills with clean recyclables, the other with waste, and trucks pick up the contents of one or the other bin and transit them accordingly to the recycling center or waste management site.

Speculation was that the system doesn’t always work that way.

The pilot began this spring with two kinds of sensors currently deployed on a total of 20 bins—half of them track recycling, the other half track waste.

In the case of recycling, the device comes with an optical sensor to identify the shapes of items as they are thrown in the bin. The optical data is sent back to the system where AI based software will differentiate a box or a bottle or a bag as well as the type of material: plastic, cardboard, paper or metal for instance. The project currently uses Wi-Fi connectivity.

The technology can even identify subtypes of plastics such as a type one or type one which may have different recycling requirements.

“We are creating this very fine-grained data to assess what’s at the bin,” Aydin said.

Tracking Waste with Organics Sensors

The devices in trash bins come with sensors for measurements of gas, temperature and humidity. That sensor data provides an understanding of food waste and food deterioration, detecting if there is a large amount of food waste, rapidly rotting, that requires an early pickup to reduce greenhouse gases.

With the waste sensors, Aydin said, researchers can use gases emitted to determine whether and how much bacterial deterioration is present, and can calculate how much waste among the households (in specific neighborhoods) is turning up in bins. Without the sensors, he pointed out, “we cannot see through the thick waste bags, to understand what is inside and how soon it needs to be collected.

Although both devices use different sensors and can employ different communication protocols, both are designed to be agnostic to the type of the bin. “We are not modifying the bins in any shape or form, they are plug and play and they are self-powered,” said Aydin.

Expanding Pilot Program

During the prototype phase, researchers built numerous types of models before selecting the bin system that is now in use.

The team now is scaling up by building casings and installing sensors into those casings which will be inserted in bins. By mid-July, the group will be installing 100 of these devices in St. Louis neighborhoods for continuous monitoring.

The research team is also preparing to build its own sensor circuits in-house and manufacture everything but the chips themselves. They then expect to be able to scale up as a research group to eventually provide large numbers of devices for larger projects, beyond St. Louis.

Results to Help Educate Residents

The data from the sensors over time is intended to help the city understand trends among neighborhoods, sociodemographic characteristics, and the effective ways of outreach and recycling education that lead to long lasting behavioral change.

Already the team has found that there are specific patterns, such as some bins consistently having good recyclables while others may have more mixed results.

Anecdotally, Aydin added, “looking at all this data one common thing I’m seeing is people tend to bag recyclables which is not something that they’re supposed to do and that’s a very common pattern.”

The devices are solar powered and store power from the sunlight hours to continue sending data during the busy time of day. Researchers are finding that the bins are mostly idle during the day when everybody’s at work, and peak active time is between 4:00 and 9:00 PM after the sun has gone down.

Preventing Cross Contamination

The data may help improve the city’s waste management program. Officials will be able to use the data to more effectively schedule waste and recycling pick up based on the levels inside bins. It may also be able to prevent cross contamination that results from mixing of clean recyclables with waste by drivers.

“Once we have an idea of where the good recycling bins are, we will be able to connect the dots to optimize the logistics of collection so that a driver will basically visit the nearest bins but also bins that contain the highest level of recyclable items in them,” Aydin said.

Another feature of the pilot is the ability to detect patterns of behavior such as illegal dumping that could lead to overflowing of one bin, which then prompts residents to throw their household waste in the only available bin.

One Tool to Minimize Flow of Plastics

If the IoT and AI technology provide the intelligence the researchers hope for, they expect it to act as a neural and nervous system of waste. Aydin likens the system to the introduction of Fitbits to step tracking and fitness. Before such trackers were available, people had little knowledge of their activity levels and behavior. With the data, they are able to begin improving their own activity levels for better health.

Ultimately, Aydin said, his goal has not changed: to minimize the amount of plastics that end up in the rivers, waterways and soil. He expects the AI-CHESS lab’s system to help identify and reduce those plastics in the natural environment.

“This may be one tool out of many,” he said, and should not preclude other efforts to tackle the plastic waste challenge.  “Remediation work is still very important, because even if we close the tap off now there are so many plastics in our rivers that on their way to our oceans it’s going to be decades of very hard work to get the plastics out of our environment.”

Learn More:

About the Author: Claire Swedberg