Enhancing UAV Systems via Task Offloading at the EDGE

Mobile Edge Computing (MEC) is a pivotal driver of 5G and subsequent mobile cellular networks, enhancing various life aspects through advanced communication and computation. MEC evaluates the computational tasks and ensures ultra-low latency as well as higher bandwidths. It has a wide range of applications, such as mobile health, surveillance systems, road infrastructure sector, and smart factory setups. A notable application is MEC-assisted autonomous navigation, where camera-fitted Unmanned Aerial Vehicles (UAVs) send images to the MEC for object detection and use the inference for controlling the UAV navigation. The offloading prolongs UAV battery life and flight duration. However, the onboard computation reduces both the battery life and flight duration, which disrupts the applications. In this study, we offload the computational tasks to both the MEC and the Cloud infrastructures in order to preserve the battery life and prolong the flight time. We present the time required for object detection inference using both the EDGE and the Cloud in latency-sensitive scenarios. The results show that employing the EDGE could outperform the Cloud both in terms of latency and throughput. We also investigate the energy consumed by both CPU and GPU that are employed with the EDGE for object detection tasks.