ADVANCING SUSTAINABLE WASTE MANAGEMENT ON CAMPUS THROUGH AN INTELLIGENT REVERSE VENDING MACHINE

Abstract

Global plastic pollution poses a critical environmental threat, impacting Sustainable Development Goals (SDGs) related to responsible consumption and climate action. Inconsistent recycling habits, particularly within high-consumption environments like university campuses, significantly contribute to this issue. Single-use plastic bottles are a major campus waste component, often ending up in landfills, exacerbating environmental degradation. This highlights an urgent need for innovative, engaging recycling solutions within educational settings and beyond. To address this, we developed an AI-driven Reverse Vending Machine (RVM) prototype and a cross-platform mobile application. The RVM utilizes a Raspberry Pi 4 as its central control unit, integrating capacitive proximity, photoelectric, ultrasonic, and infrared sensors for comprehensive bottle detection and monitoring. The system employs a YOLOv5 model for object detection, trained on a robust dataset of 2,000 labelled images, and a Flutter-based mobile application for user interaction and reward redemption. The development followed an Agile methodology, emphasizing iterative testing and refinement. A campus pilot study confirmed the system's efficacy. The YOLOv5 model achieved high accuracy (99.0% mAP@0.5, 98.4% precision, 96.8% recall). The system was highly responsive, with a detection-to-reward cycle under 1.2 seconds. User feedback was positive (SUS score: 81.3), and a student survey showed 76% willingness to use the RVM, with 48% motivated by vouchers, indicating strong acceptance and potential for behavioral change. This AI-driven RVM offers a technically feasible and highly accepted solution for sustainable waste management, providing an effective, user-centric approach to combat plastic pollution and offering a blueprint for broader global recycling efforts.