Robust Symbol Detection in Large-Scale Overloaded NOMA Systems

We a framework for the design of low-complexity and high-performance receivers multidimensional overloaded non-orthogonal multiple access (NOMA) systems. The is built upon novel compressed sensing (CS) regularized maximum likelihood (ML) formulation discrete-input detection problem, in which l 0 -norm introduced to enforce adherence solution prescribed discrete symbol constellation. Unlike much preceding literature,.g., (Assaf et al., 2020, Yeom 2019, Nagahara, 2015, Naderpour Bizaki, Hayakawa Hayashi, 2017, 2018, Zeng 2020), method not relaxed into xmlns:xlink="http://www.w3.org/1999/xlink">1 -norm, but rather approximated with continuous asymptotically exact expression without resorting parallel interference cancellation (PIC). objective function resulting thus sum concave-over-convex ratios, then tightly convexified via quadratic transform (QT), such that its can be obtained iteration simple closed-form closely resembles classic zero-forcing (ZF) receiver, making particularly suitable large-scale set-ups. By further transforming aforementioned problem quadratically constrained program one convex constraint (QCQP-1), optimal regularization parameter used at each step iterative algorithm shown largest generalized eigenvalue pair matrices are given closed-form. so obtained, referred as Iterative Discrete Least Square (IDLS), extended address several factors practical relevance, noisy conditions, imperfect channel state information (CSI), hardware impairments, yielding Robust IDLS algorithm. Simulation results show proposed art significantly outperforms both receivers, linear minimum mean square error (LMMSE), recent CS-based state-of-the-art (SotA) alternatives, sum-of-absolute-values (SOAV) complex sparse regularizers (SCSR) detectors. It also simulations technique integrated existing detection-and-decoding (IDD) methods, accelerated convergence.