Deep Learning Based Channel Estimation for UAVs: A Modified U-Net Approach

stable and reliable communication link is crucial for unmanned aerial vehicle (UAV) applications. Key challenges include the UAV's high mobility (10-100 km/h) and an unstable data link. Orthogonal frequency division multiplexing (OFDM) enables higher data rate transmission with improved bandwidth efficiency, while minimizing channel effects on the received signal and enhancing bit error rate (BER) performance. This article proposes a deep learning based channel estimation (CE) for 802.11ac OFDM systems considering the mobility of the receiver. The proposed CE algorithm is a two-step process. The first step uses an especially developed deep neural network built on the U-Net model for denoising the signal received, followed by least squares (LS) estimation in the next step. The simulation results show that the proposed model has improved the BER by 50% and 40%, the data rate by 10% and 7% and outage probability by 10% and 7%, Arespectively, when compared to the conventional LS estimator and machine learning based LS estimator. The proposed model has also been evaluated for three different modulation schemes, i.e., QPSK, 16-QAM, and 64-QAM and the complexity analysis has been done to strengthen our studies further.