AI increases efficiency in waste management

Despite a push for circularity in our industrial systems, regulations such as Extended producer responsibility (EPR) for packaging and supply-side shocks, such as access to raw materials and disrupted supply chains because of COVID-19 and geopolitical conflict, we still have a global problem with waste going to landfill or into the environment. 

Waste generation continues to grow and is predicted to reach 3.4bn tonnes by 2050. Rising levels of waste are also linked to the rapid growth in consumption. The Circularity Gap report 2024 notes that more than 90% of the materials extracted and used to make products and packaging ultimately become waste. In essence, only 7.2% of these materials go back into the production system.

This is poor from both resource efficiency and climate perspectives. For example, according to the UN International Resource Panel, the production of materials accounted for 23% of global carbon emissions in 2015, although this was up from 15% in 1995.

While the media has focused on household and consumer waste, the volume of industrial waste generation globally is 18 times greater, according to the World Economic Forum's 2021 article. The infrastructure sector is responsible for 30% of this, but next on the list – and growing rapidly – is material generated by manufacturers. Although global manufacturing output is more than $40tr per year, 20% of every dollar spent in manufacturing is wasted due to inefficiencies.

In particular, the waste and recycling sector has adopted sensor systems that measure the volume of material in containers and vehicles – the so-called internet of bins rather than internet of things.

For example, the City of Edinburgh Council has installed 11,000 such sensors across its on-street bins to monitor and optimise container deployment and collection. Sensors or smart labels printed onto packaging can also be embedded in reusable packaging and goods, in a substrate enabling the asset to be tracked.

Meanwhile, the waste and recycling sector has deployed AI in the form of machine vision systems for robotic sorting of packaging waste in materials recovery facilities. High-resolution cameras above conveyor belts linked to AI-driven software identify the different types of material used in packaging, which can then be sorted. The cameras can also identify brand names, so companies have a better idea of how much of their packaging consumers recycle.

At a micro scale, machine vision and weighing are also being used in commercial real estate to measure recycling and general waste. Hospitality operators are likewise using such systems to measure and reduce food waste.

Reusable systems include for instance, coffee cups with an embedded sensor, which can be returned directly to the retailer or through a reverse vending machine, where they are collected, washed and returned for reuse.

In September, the European Parliament also approved new measures to prevent waste food and textiles, which will require digital technologies to measure and track material through sale and recovery. Textile producers for instance will have to sign up to an EPR programme, requiring them to cover the cost of collection, sorting and recycling of textile products.

Meanwhile, the UK Department for Environment, Food and Rural Affairs is set to roll out a mandatory digital waste tracking system next spring. After various iterations, this will require all waste and recycling operators to use approved software that can readily export data for each nation's environmental regulator.

At the same time, England, Scotland and Northern Ireland will see the roll-out of a deposit return scheme for single-use plastic and metal drinks containers, with beverage producers adding a refundable deposit to the cost of a product.

This will be redeemable using a network of reverse vending machines that will read the smart labels on each returned container. The aim is to reduce littering, improve recovery rates and generate quality recycled material by offering a financial incentive to return the item.

For instance, bin-fill sensors see material put into on-street bins, but not where this goes after the bin is emptied. Machine vision in turn only sees material coming into and through material recovery facilities, and does not track it beyond that site. Bin sensor systems do not qualify the type of material either, and machine vision systems only measure what they are trained to identify.

The challenge for the global urban and industrial waste system is that the pathways for all this material – recyclable, general or hazardous – are complex. The generation, movement and processing of waste is a dynamic system of material flowing around organisations and territories.

These materials emerge from operational sites and are then collected by haulers or carriers and taken to locations often operated by other firms, where the material may be sorted, processed and moved to further sites. What happens to material will be determined not only by where in the world the operators are based but by the amount of waste, its value, the costs, the haulers and contractors, the infrastructure, the regulations and a host of other factors.

In the next five years, sensor, data and robotic systems will be more widely deployed, underpinning a more efficient and effective materials system and supporting circular business models. These systems will be deployed by city and national governments, waste and recycling operators and the organisations that produce waste.

Through judicious and widespread digitisation, it is possible to build digital twins of complex, global value-chains that can enable the systemic changes needed to reduce waste at source and maximise the recovery of material.

The very complexity of the datasets involved lends itself to digitisation and AI to automate, process and analyse and create a clear view of the system. While technology can be the enabler, growing circularity will also need regulation, investment and a realignment of priorities.

While there are compliance schemes and systems for classifying waste, because of that waste's complexity and volume there is significant variation in the way it is measured. At the same time, companies that generate such waste are realising that it is time to rethink their approach. This has partly been prompted by the need to convert raw materials more efficiently, by increasing pressure to take responsibility for waste, and by the recognition that it has a market value that is not being realised.