Enhanced NOMA System Using Adaptive Coding and Modulation Based on LSTM Neural Network Channel Estimation

Abstract

Non-orthogonal multiple access (NOMA) is the technique proposed for multiple access in the fifth generation (5G) cellular network. In NOMA, different users are allocated different power levels and are served using the same time/frequency resource blocks (RBs). The main challenges in existing NOMA systems are the limited channel feedback and the difficulty of merging it with advanced adaptive coding and modulation schemes. Unlike formerly proposed solutions, in this paper, we propose an effective channel estimation (CE) algorithm based on the long-short term memory (LSTM) neural network. The LSTM has the advantage of adapting dynamically to the behavior of the fluctuating channel state. On average, the use of LSTM results in a 10% lower outage probability and a 37% increase in the user sum rate as well as a maximal reduction in the bit error rate (BER) of 50% in comparison to the conventional NOMA system. Furthermore, we propose a novel power coefficient allocation algorithm based on binomial distribution and Pascal’s triangle. This algorithm is used to divide power among N users according to each user’s channel condition. In addition, we introduce adaptive code rates and rotated constellations with cyclic Q-delay in the quadri-phase shift keying (QPSK) and quadrature amplitude modulation (QAM) modulators. This modified modulation scheme overcomes channel fading effects and helps to restore the transmitted sequences with fewer errors. In addition to the initial LSTM stage, the added adaptive coding and modulation stages result in a 73% improvement in the BER in comparison to the conventional NOMA system.

Transceiver Design for Full-Duplex UAV Based Zero-Padded OFDM System With Physical Layer Security

“…The authors of [32] investigated the performance of the NOMA system considering outage probability, capacity, BER and user power allocation accuracy. In their proposed system, long-short term memory (LSTM) neural network with adaptive channel coding and digital modulation was introduced.…”

A Novel Spread Spectrum and Clustering Mixed Approach with Network Coding for Enhanced Narrowband IoT (NB-IoT) Scalability

“…This orthogonal approach is also referred to as overlay spectrum sharing as opposed to the underlay (non-orthogonal Multiple Access (NOMA)) spectrum sharing [ 18 ]. In the overlay approach, different users (Cellular User Equipments (UEs), as well as IoT (NB-IoT and LTE-M) nodes, are allocated separate time or frequency resource blocks (RBs) with the same power level, while in the underlay approaches, different users are allocated different power levels but are serviced using the same time or frequency resource blocks [ 19 ]. The difference between the two approaches is illustrated in Figure 1 .…”

Successive interference cancelation‐inspired channel estimation for downlink non‐orthogonal multiple access

“…There are numerous accounts of CE, specific to downlink (DL) NOMA that have been studied. [45][46][47][48][49][50][51] The work in Reference 45 proposes a CE algorithm based on long-short term memory neural network. Performance improvements over conventional NOMA systems are reported in Reference 45, and they are attributed to the adaptability of the approach to channel state fluctuations.…”

“…Most of the work in the literature that focus on CE for DL NOMA systems do not characterize their CE performance evaluations using the mean squared error (MSE) as the metric for analysis. Furthermore, papers [43][44][45][46] consider a scenario whereby all pilots in the superimposed signal are known at each receiver. The requirement of known pilots is relaxed in this paper by employing partial knowledge of the pilots within the superimposed signal.…”