Hardware for a Circular Planet
An open-source blueprint for automated, high-accuracy urban waste sorting. Built from the very electronics it was designed to rescue.
An open-source blueprint for automated, high-accuracy urban waste sorting. Built from the very electronics it was designed to rescue.
Global waste management is failing at the most critical stage: **Sorting**. While high-discipline communities like Osaki or Kamikatsu achieve near-zero waste through 45-category sorting, urban populations lack the time and infrastructure to replicate this manually.
This project proposes a **COTS-based automated micro-station** to solve the "last-mile" recycling problem. By removing the complexity from the user and giving it to AI, we can achieve high-purity recycling streams for materials that currently end up in landfills.
Our algorithm uses a multi-layered sensor strategy to identify and categorize dry waste with surgical precision. Each category uses a unique logic path derived from salvaged electronic sensors.
Utilizes Computer Vision to identify model-specific geometry. Logic then activates **Inductive Proximity Sensors** to detect high-density PCB copper and scans for **Passive RFID/NFC** identifiers embedded in modern devices.
Distinguishes ferrous and non-ferrous alloys using **Eddy Current Analysis**. Neodymium magnets (salvaged from 3.5" HDDs) identify iron/steel, while non-ferrous alloys like Aluminum are identified via high-frequency resonance checks.
Goes beyond bottle glass. The station identifies Borosilicate vs. Soda-lime by measuring the **Refractive Index** using salvaged optical sensors from flatbed scanners and targeted light transmission tests.
AI identifies resin types (PET, HDPE, PVC) by analyzing texture and spectral reflection. Multi-material items are flagged for human-in-the-loop verification or AGI classification.
Focuses on fiber weave recognition. Using macro-photography sensors, the system distinguishes between organic fibers (Cotton, Wool) and synthetic polymers (Polyester, Nylon) to ensure textile circularity.
Determines fiber length and moisture levels. Logic prevents contamination by identifying non-recyclable coated papers (wax/plastic coatings) using surface gloss analysis.
The station proves the circular economy by using salvaged **Commercial Off-The-Shelf (COTS)** hardware as its primary sensory organs and brain.
A simple interface hiding complex machine-learning decisions. Our algorithm ensures a friction-less experience for the public while maintaining high-purity outputs.
User approaches and places dry waste on the scanning table. The station identifies the user via NFC or App for gamification rewards.
Cameras and sensors perform simultaneous checks. The AI model compares data against a massive library of recyclable waste examples.
The station guides the user or automatically opens the specific bin (e.g., Industrial Electronics, Glass, Textiles).
