<![CDATA[While Smart City initiatives are evolving, waste management infrastructure remains a critical bottleneck, often hindered by high energy dependency and latency issues associated with cloud computing. Traditional automated solutions lack the autonomy required for scalable, outdoor deployment. This research introduces Orange Box a self-sustaining Edge-AI waste classifier designed to bridge the gap between high-performance computing and energy efficiency. The primary goal is to demonstrate that complex Deep Learning tasks can be executed locally on renewable energy without sacrificing classification precision. The system orchestrates a MobileNetV2 architecture on the Raspberry Pi 5, utilizing TensorFlow Lite (TFLite) quantization to drastically reduce computational load. Uniquely, this Green IoT node is fully decoupled from the power grid, driven by a custom power management system utilizing a 100Wp monocrystalline solar panel to sustain both the neural processing unit and robotic actuators. Experimental benchmarks reveal a robust 92% classification accuracy with an inference latency of just 45ms, significantly outperforming previous edge-device generations. Crucially, energy analysis validates operational autonomy for up to 72 hours without sunlight, confirming the system’s reliability for continuous urban deployment. This study demonstrates that the convergence of quantized Edge AI and solar harvesting is not merely theoretical but a deployable standard for the next generation of Smart City infrastructure, directly advancing the Sustainable Development Goals (SDGs) for sustainable urbanization.]]>
Copyrights © 2026
