1. General
Category
SDG 8: Decent Work and Economic Growth
SDG 11: Sustainable Cities and Communities
SDG 13: Climate Action
Category
Energy & Natural Resources
2. Project Details
Company or Institution
AMP Robotics
Project
AMP Robotics' AI platform and capabilities
General description of the AI solution
AMP Robotics’ proprietary AI platform, AMP Neuron™, encompasses the largest known real-world dataset of recyclable materials, with the ability to classify more than 50 different categories of recyclables across single-stream recycling, e-scrap, and construction and demolition debris with an object recognition run rate of more than 10 billion items annually.
This combination of scalable accuracy and classification creates a step-level solution for data collection and measurement that materials recovery facilities can use to optimize their operations; reclaimers, mills, and manufacturers can use to validate that incoming feedstock meets specifications and standards for chemically compliant bales; and brand owners and government stakeholders can use to measure the quality, flow, and recovery of recyclable materials.
Neuron applies computer vision and deep learning to guide the company’s robots to precisely identify, differentiate, and capture recyclables found in the waste stream by color, size, shape, opacity, form factor, and more, storing data about each item it perceives. It can recognize and recover material as small as a bottlecap and as unique as a Keurig coffee pod from complex material streams. It can quickly adapt to container packaging introduced into the recycling stream with recognition capabilities to the brand level—increasingly critical as demand for sufficient quantities of high-quality recycled material grows to meet consumer packaged goods companies’ commitment to use of post-consumer recycled content. And the more robots and sensors that AMP deploys into production, the more a network effect is created, meaning the company is able to exponentially increase its sorting intelligence by installing more units.
Website
Organisation
AMP Robotics
3. Aspects
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
The AMP Neuron™ AI and computer vision system uses machine learning to continuously train itself by processing millions of material images into data, building upon an ever-expanding neural network that adapts to changes in a facility’s material stream. AMP’s AI technology works by looking, through a hyperspectral camera, at images of conveyor belts within recycling facilities. To train the AI, AMP shows examples of materials that belong to each category, and dozens of data annotators comb through these images and classify objects belonging to each material type. This process is managed by a custom tool built to ensure exceptional accuracy, and allow for various team members to simultaneously contribute to the creation of a dataset. The company’s millions of labeled images drive its highly accurate image classification models, capable of identifying thousands of recyclables in milliseconds, so its robotic arms can sweep into action. As this data set takes shape, the examples feed AMP’s proprietary machine learning algorithms. Training with a massive dataset from packaging types from all over the world, different lighting conditions, and many unique recycling businesses has provided AMP as much as 99 percent accuracy on inference.
AMP’s approach to R&D is based on swift, iterative innovation. With more than one dozen patents or patents pending, the company is developing and enhancing new AI and robotics applications for recycling, from material-specific feature sets to performance enhancements that support more precise automated sortation and rapid recovery of high-value commodities. AMP has received accolades from the likes of Fast Company, Fortune, Grist, and more for its spirit of innovation.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
There’s a huge demand for recycled content, and recycling provides a stable source for this demand. Recycling helps reduce greenhouse gas emissions by lowering energy consumption, and using recycled materials to make new products reduces the need for virgin materials. Since inception, AMP’s technology has eliminated 157,086 metric tons of greenhouse gas emissions, equating to taking 32,726 cars off the road. Thus, AMP’s AI-driven recycling process supports the SDG of climate action.
The Institute of Scrap Recycling Industries (ISRI) has compiled statistics on how recycling supports jobs and job creation, in alignment with the SDG of decent work and economic growth:
164,154 jobs supported by the recycling and brokerage operations of the scrap industry in the U.S.
367,356 jobs supported by the industry through suppliers and the indirect impact of the industry’s expenditures
Thousands of indirect jobs of people in other sectors such as servers in restaurants, construction workers, teachers, and other professionals.
Moreover, in addition to job support and creation, recycling technology helps remove humans from dull, dirty, dangerous jobs and offers a path to higher-skilled, more desirable roles in maintaining and managing the technology. Because the recycling industry has such a hard time filling these roles in the first place, technology like AMP’s help them run fully staffed in order to economically recover more material.
AMP’s pursuit of secondary sortation driven by AI also advances the SDG of sustainable cities and communities. Secondary MRFs offer the flexibility to sort the complex commodities, and those with low critical mass in the material stream. By addressing these legacy obstacles to the expansion of recycling, AMP’s technology is helping to increase recycling rates to improve the world we collectively inhabit by broadening modern recycling infrastructure and access to recycling in historically underserved communities and geographies.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
AMP has deployed more than 150 systems globally. They’re primarily located in North America, but the company also has systems in Japan and is expanding into Europe. AMP’s customer base grew from early adopters who were partners to the company in testing the technology, and now includes municipalities, privately owned recycling businesses, and large waste management companies like Waste Management and Waste Connections.
COVID forced many recycling businesses to suspend operations due to concerns for worker safety, as facilities weren't designed with social distancing in mind. Simultaneously, the pandemic increased demand for high-quality recycled feedstock to overcome supply chain interruptions and shifts in raw material availability. The company saw an uptick in new orders and repeat orders from existing customers. Customers were also more willing to adopt the technology more aggressively, and there were instances where they were buying more systems versus starting with just one or two.
Since inception, AMP has eliminated 157,086 metric tons of greenhouse gas emissions, equating to taking 32,726 cars off the road. Over a 12-month period, AMP’s AI-driven fleet identified and sorted one billion pieces of material. As it scales both geographically and across the recycling value chain, AMP is pursuing expanded applications of AI like secondary sortation. AMP’s application of AI for material identification and advanced automation has matured to the point where it’s become feasible to develop secondary sortation facilities, economical to deploy and sustain nationally, that process low volumes of difficult-to-recycle mixed plastics, paper, and metals sourced from residue supplied by primary MRFs. With the U.S. recycling less than 33% of recyclable materials overall—and just 9% of plastics—produced annually, the ability to recover recyclables from residual waste streams represents a major opportunity to increase national recycling rates.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
The current MRF infrastructure is insufficient to capture the balance of recyclables lost to landfill annually. Legacy methods of primary material recovery are designed to handle high-volume recyclables, limited to a narrow range of material types and form factors, which results in tons of residual material leaking to landfill. This is primarily dictated by stronger markets for #1 and #2 plastics, and possibly #5 plastics at facilities with high throughput. Even with a rise in prices, the switching costs to obtain a critical mass of increasingly complex packaging restricts MRFs from taking action. MRFs are being pushed to limit the materials they sort. The high-cost burden of the industry’s current infrastructure compounds this problem, making recycling economically unfeasible in a number of geographies. The net result is billions of tons of recyclables and billions of dollars lost to landfill despite the demand for recycled content.
AMP’s pursuit of secondary sortation driven by AI establishes secondary MRFs with the flexibility to sort the complex commodities, and those with low critical mass in the material stream. By addressing these legacy obstacles to the expansion of recycling, AMP’s technology is helping to increase recycling rates to improve the world we collectively inhabit by broadening modern recycling infrastructure and access to recycling in historically underserved communities and geographies.