Air–Ground Surveillance Sensor Network based on edge computing for target tracking

Abstract

With the development of Internet of Things (IoT), Unmanned Aerial Vehicle (UAV) target tracking recently has received lots of attention in research community. The two major research topics in UAV target tracking include loss of a tracked target and high tracking latency, especially in the case where tracking is conducted in an area of many sight-blocking obstacles that put restrictions on a UAV’s flight altitude and block its line of sight. Most recent work on these topics focuses on improving tracking accuracy through adjusting tracking algorithms based on deep learning. But the complexity of these algorithms requires a computation capacity that a regular UAV cannot afford, and therefore tracking failure probability is still non-negligible. To address this challenge, we propose a new target tracking system, referred to as an Air–Ground Surveillance Sensor Network (AGSSN). We build AGSSN by jointly optimizing the network establishment and data transmission, and we design an algorithm ARIT to achieve an optimal tracking performance. We further carry out a series of simulations by deploying our AGSSN on a university’s campus map, and our simulation results show that our proposed AGSSN system can achieve higher reliability and significantly better performance than regular tracking systems in an area with many visually blocking obstacles.