Work

Deep learning for multiuser detection in UAV communications using RIS and orthogonal time sequency multiplexing with circular pulse shaping

Year: 2026
Type: article
Abstract: This paper addresses reliable multiuser signal detection and perceptual-level physical-layer confidentiality for circular pulse-shaped orthogonal time-sequency multiplexing (OTSM) transmission over reconfigurable intelligent surface (RIS)-assisted unmanned aerial vehicle (UAV) delay–Doppler channels. To improve multiuser multiple-input multiple-output (MIMO) detection under doubly dispersive propagation, a feedforward deep neural network (DNN) trained using 12-fold cross-validation is employed. The proposed learning-based detector achieves lower mean square error (MSE) than the considered minimum mean square error (MMSE), zero-forcing (ZF), orthogonal approximate message passing (OAMP), and OAMP-Net baselines, and provides improved bit error rate (BER) performance particularly at moderate and high signal-to-noise ratio (SNR) levels, while maintaining stable convergence and validation behavior across users. To provide lightweight physical-layer confidentiality, a nine-dimensional (9D) eighth-order chaotic encryption mechanism with random signal insertion is incorporated into the transmission chain, and perceptual confidentiality is evaluated using the Learned Perceptual Image Patch Similarity (LPIPS) metric. Block diagonalization (BD) precoding is used to mitigate multiuser interference, while repeat–accumulate (RA) coding improves detection reliability. Simulation results show that the proposed framework achieves 30–40 dB out-of-band (OOB) power reduction and a peak-to-average power ratio (PAPR) of approximately 11 dB at a complementary cumulative distribution function (CCDF) of \(10^{-3}\) . Under the idealized simulation setting with perfect channel state information (CSI), BD-based interference suppression, and no pilot or RIS-control overhead, the system attains an upper-bound aggregate block-normalized spectral efficiency of approximately 227 bps/Hz. This value is reported as a system-level benchmark for the considered model and should not be interpreted as a practical per-subcarrier or per-antenna spectral efficiency. Encrypted image transmission yields LPIPS values above 1.19 for all users, indicating strong perceptual dissimilarity prior to reception. Overall, the results indicate that the proposed RIS–UAV-assisted circular pulse-shaped OTSM framework can improve spectral containment, signal detection reliability, and perceptual-level physical-layer confidentiality in highly dynamic wireless environments.
Language: English
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7234965392
FWCI:
Cites: 54
Cited by:
Related to: 10