A single dataset can be used in different ways by different people. Waste data is no exception.
Sorting machinery manufacturers, developers of recovery facility software, plant managers and group-level strategists all find different uses — and different insights — in the data generated by AI for waste management.
It’s why we’ve focused on making that information as accessible and flexible as possible with integrations. We believe that waste data can enrich the entire waste ecosystem, rather than a single process or brand of machinery.
Here’s how our Application Programming Interface (API) works, and why we’ve created different integrations to cater to specific use-cases.
The right speed for the right task
For those without a background in development, an API stands for Application Programming Interface and is the method used to allow one tool (like an AI monitoring platform) to share information with another (like a sorting robot). The process “integrates” one tool with another, hence the term “integrations”.
Latency is the time it takes for information to be shared. When time is of the essence, latency needs to be incredibly low. In other scenarios, we sacrifice some speed in favour of granular detail.
Integrations aren’t one-size-fits-all. We’ve created APIs that work at different speeds, depending on how people want to use Greyparrot data. In some use-cases, every millisecond counts:
Milliseconds: AI-powered sorting arms
When it comes to real-time sorting, speed really does matter. Robotic arms only have a second or two to identify, pick, and dispose of waste objects on a moving conveyor belt.
Belts can be slowed to improve their sorting accuracy, but that means sacrificing the amount of material you’re recovering (also known as throughput).
AI-powered reflexes
You’re more likely to catch a ball if you can spot it in time to move your hands. The robotic API operates a similar principle: the faster a robot can accurately spot a contaminant, the better. Our AI does the spotting, and shares the object’s location so the robot can remove it.
Thanks to close collaboration with robotics leaders like GEKU and ABB, we’ve cut the time it takes for our AI to detect the location of a waste object from 250 milliseconds to just 25. It’s an exponential improvement, and an incredibly fast integration that can be used to make any brand’s sorting machinery more accurate.
Seconds: Processes that adapt with solid waste composition
Not all use-cases require instant updates. When optimising processes or machinery, it’s more valuable to create a feedback loop that measures output and adjusts accordingly. Material composition doesn’t change every millisecond, but it can change by the second.
Automatic improvements
Aided by our robotic API, we’re working with sorting machinery providers to build mechanical separators that auto-adjust their settings as material composition changes:
Our units gather data at the input and output of their machinery, tracking composition at both stages. As composition changes or efficiency drops, the separator automatically adjusts its parameters in response.
Minutes: Data-backed recovery facility improvements
Hindsight is 20-20: when looking for trends and patterns in material flows, detailed historical data offers invaluable perspective. Collaborators like Sortflow integrate with our cloud API to offer insight on sorting performance, while major plant builders can use it to model future plant designs, and identify areas for infrastructure investment.
Data-backed investment
Each of our monitoring units gathers hundreds of millions of data points every year, making it possible to analyse material from specific date ranges on a minute-by-minute basis.
Searching through massive datasets to gather that information from the cloud takes time, though. There’s still balance to be found: we’ve structured our database to reduce lookup times, making it possible to create reports in platforms like Tableau and PowerBI in minutes or less.
Improving access to AI for waste management
Whether insight is needed every millisecond or every few months, we’re working to make our AI’s recognition capabilities as accessible as possible.
We describe it as an “agnostic” approach — in other words, the data gathered by our units isn’t limited to specific machinery or software. Instead, it’s available to power everything from robotics to reporting tools with cognitive intelligence.
By collaborating with fellow specialists and innovators, we’re creating facilities that are greater than the sum of their parts. The more tools we can power, the more efficient waste management gets.
