Combating the Impact of Jittering in UAV-based Sensing Systems Using Deep Denoising Network

In this paper, we exploit the deep learning based technologies to mitigate the impact of unmanned aerial vehicle (UAV) jittering on wireless sensing performance. In recent years, UAV has been widely utilized for remote sensing applications due to its high flexibility and maneuverability. However, the mobility and vibration of the UAV's body may cause the jittering effect which can severely degrade the sensing performance. To our best knowledge, the impact of UAV jittering has not been fully examined in literature so far. To alleviate this problem, we propose to leverage adversarial denoising autoencoder (ADAE) for corrupted signal reconstruction. To validate the effectiveness of our proposed scheme, we consider a device-free human sensing scenario in which a UAV is used to sense surrounding human activity by analyzing the received signal strength (RSS). Experiments demonstrate that the proposed ADAE based scheme can effectively reduce the impact of UAV jittering, recovering up to 97% of the performance loss due to the UAV jittering.