Joint Optimization of E2E Latency, FPS, Energy, and Confidentiality for Surveillance UAV

In this paper, we investigate an unmanned aerial vehicles (UAVs) with edge computing (EC) servers for surveillance applications. We address the challenges of object detection using deep neural networks (DNNs) under constraints of limited computational capability and energy resources on UAVs. To tackle the challenges, we propose the UAV Surveillance Optimization Framework (USOF) algorithm, by employing a Lyapunov optimization framework to jointly optimize four key performance metrics: end-to-end latency, frames per second (FPS), energy consumption, and data confidentiality during computational offloading. Numerical results verify that our proposed USOF algorithm achieves a balanced performance for surveillance UAVs.