Autonomous Waste Classification Using Multi-Agent Systems and Blockchain: A Low-Cost Intelligent Approach

Area of Science:

• Environmental Science
• Computer Science
• Artificial Intelligence

Background:

• Growing global waste generation necessitates advanced, sustainable waste management solutions.
• Current waste management systems often lack efficiency and transparency, hindering effective recycling and resource recovery.

Purpose of the Study:

• To develop and implement a prototype smart bin for automated waste sorting.
• To integrate a multi-agent system and blockchain technology for intelligent waste classification and transparent data management.

Main Methods:

• Utilized sensors for waste identification (organic, plastic, paper).
• Employed a multi-agent system for collaborative, real-time sorting decisions.
• Integrated blockchain for immutable and transparent waste registration and traceability.
• Evaluated YOLOv8, YOLOv11, and YOLOv12 for waste detection and classification, selecting YOLOv12 based on performance.

Main Results:

• YOLOv12 demonstrated superior performance with mAP@50 of 86.2%, 84.6% overall accuracy, and 80.1% F1 score.
• Achieved processing latency below 9 ms per image with YOLOv12, enabling near-real-time application.
• The smart bin prototype confirmed viability in accuracy and computational efficiency.

Conclusions:

• The developed smart bin offers an innovative, cost-effective, and scalable solution for ambient intelligence in waste management.
• This system lays the groundwork for smart waste management infrastructures in sustainable urban environments.
• The integration of AI and blockchain enhances waste sorting efficiency, traceability, and transparency.