Comprehensive Study of Energy-Aware Framework for UAV-Based AIoT Node Using Edge Intelligence

Abstract

Unmanned Aerial Vehicles (UAVs) are increasingly employed as mobile IoT nodes for smart cities, precision agriculture, and emergency response. However, the integration of AI-driven decision-making onboard UAVs is severely constrained by limited computational resources and energy availability. This paper introduces a low-power AIoT framework that enables UAVs to perform intelligent tasks under strict power and hardware limitations. The proposed architecture distributes decision-making between embedded AI modules running on STM32-based flight controller and edge/cloud servers, dynamically balancing onboard processing and task offloading. The framework employs a lightweight AI model that handles real-time control tasks locally, while computationally intensive inference is offloaded to IoT gateways when communication links and latency allow. This architecture is validated using a Hybrid simulation platform built around XPlane 12, Python middleware, STM32 microcontroller, and Nvidia Jetson Nano as an AI computing platform. Test scenarios include UAV trajectory tracking and wind disturbance rejection. Results demonstrate that the proposed system reduces onboard energy consumption while preserving high control accuracy and robustness. Furthermore, mission endurance is significantly extended compared to conventional embeddedonly approaches. This work highlights the potential of lowpower edge AI for sustainable UAV-IoT integration, opening pathways for long-endurance missions.