GANCE: Generative Adversarial Network Assisted Channel Estimation for Unmanned Aerial Vehicles Empowered 5G and Beyond Wireless Networks

Abstract

Unmanned aerial vehicles (UAVs) have discovered a plethora of societal applications such as remote sensing, disaster management, medical emergency, security and surveillance, etc. UAVs require fast and reliable communication lines, and selecting the appropriate channel estimation (CE) technique is pivotal in reliable communication. Orthogonal time frequency space (OTFS) is an innovative modulation technique designed to support reliable communication in rapid mobility environments for 5G and beyond applications, effectively addressing challenges posed by Doppler shifts and multipath propagation. Current OTFS receivers utilize threshold methods such as least squares (LS) and minimum mean square error estimators for CE. To further enhance the accuracy and robustness of the CE process, this paper proposes a generative adversarial network (GAN) for learning channel parameters and performing CE in OTFS-based communication systems for high-speed UAVs (100-500 km/h). Firstly, a system model considering the Doppler effect has been modeled mathematically, and then the solution to the CE problem is presented for UAVs-assisted networks using GAN. The proposed GAN architectures comprise a U-Net-based generator and a PatchGAN discriminator for adversarial training of the model. The proposed model is compared with the baseline approaches in terms of bit error rate (BER), outage probability (OP), and normalized mean squared error (NMSE) for different velocities and modulation schemes. The proposed model has given an improvement of 70%, 55%, and 45% in BER performance and 40%, 30%, and 20% in OP compared to the conventional LS estimator, machine learning-based estimator, and deep learning-based estimator, respectively. The proposed model has also demonstrated robustness against the Doppler effect caused by the high mobility of UAVs, with only a minimal decrease in NMSE performance of 2-3 dB for every 50 km/h increase in speed. Additionally, the time complexity and processing time of the proposed model have been studied to test its suitability for UAV applications.

Keywords

• Unmanned aerial vehicles
• channel estimation
• orthogonal time-frequency space
• generative adversarial network
• U-Net
• deep learning