<![CDATA[Traditional waste bins provide only passive storage and rely on rigid collection schedules, which often causes overflow, premature pickups, and mixed hazardous (B3) waste that increases health and environmental risks. This study proposes the Source-Level Waste Intelligence (SLWI) architecture, which coordinates source-level segregation, condition awareness, spatial awareness, and rule-based decision-making at the point of disposal. The architecture is implemented through the BINTAR smart waste-bin prototype, integrating an EfficientNetB0 image-classification model with servo-driven sorting, HC-SR04 ultrasonic and MQ-135 gas sensing, a SIM808 GPS module, and a web-based ReactJS monitoring dashboard on an ESP32 controller. Functional testing shows that the EfficientNetB0 classification model achieved 96.99% test accuracy, the ultrasonic sensors produced a mean absolute error of 10.33 cm, the MQ-135 sensor produced distinguishable responses across predefined gas/odor response ranges, and the SIM808 module successfully acquired GPS coordinates in all documented scenarios. Website updates were observed within 1–5 s under the prototype test conditions. These results demonstrate the functional integration of segregation, condition sensing, and geotracking within a single coordinated architecture, while quantitative physical-sorting accuracy and positioning error remain subjects for future validation.]]>
Copyrights © 2026
