Sharif: Edge Computing and Energy Consumption Optimization for UAVs at Various Altitudes

Unmanned Aerial Vehicles (UAVs) are essential for enhancing situational awareness and mitigating risks in mission areas prone to threats. However, integrating deep learning architectures into UAVs is challenging due to payload and energy limitations, which can cause rapid battery depletion and reduced operational duration, limiting their effectiveness in extended missions. This paper presents a novel electronic architecture for UAVs to optimize energy consumption during UAV missions while maintaining high performance. The architecture dynamically switches between two accelerators, a Visual Processing Unit (VPU) and a Graphics Processing Unit (GPU), based on real-time mission requirements, enabling seamless transitions and efficient handling of both simple and complex tasks. Experimental evaluations show that this approach reduces energy consumption by up to 26% compared to traditional systems, thereby extending UAV operational endurance and ensuring high service quality across multiple altitudes.