A Trade-off Analysis of Latency, Accuracy, and Energy in Task Offloading Strategies for UAVs

Object detection is a critical aspect of computer vision, particularly for unmanned aerial vehicles, commonly known as drones. Drones depend on object detection to perform various tasks such as surveillance or search and rescue. However, drones face challenges such as limited computation and small object detection, which requires high accuracy within tight timeframes. The emergence of edge computing allows real-time processing, balancing accuracy and latency by offloading tasks to more powerful edge servers. However, efficient resource allocation remains a challenge. This paper presents results from real-world experiments that explore the trade-offs among latency, accuracy, and energy consumption in UAV object detection. Notably, the experiments demonstrate that offloading can reduce latency, particularly for tasks demanding high levels of accuracy. These results can inform offloading and resource allocation decisions.