SCRAM: A Novel NOMA-Based Scheme for Reliable Ultra-Low Latency 6G Applications

In this thesis, a novel scheme dubbed SCRAM, which stands for Slotted Coded Random Access Multiplexing, is presented. The proposed SCRAM scheme is provisioned to meet the challenging requirements of the future 6G IoT applications, portrayed in massive connectivity, reliability, and ultra-low latency. SCRAM is set to operate in highly overloaded scenarios, which places it under the umbrella of Code-Domain Non-Orthogonal Multiple Access (CD-NOMA). Moreover, in order to meet the low latency requirement, SCRAM incorporates Slotted ALOHA as a candidate random access scheme. However, the uncoordinated access leads to collisions among the transmitted packets of the various users. In order to provide reliability, the proposed SCRAM tends to protect the transmitted symbols by deploying powerful capacity-approaching LDPC codes. The essence of the proposed SCRAM lies in its novel joint iterative decoding structure. Unlike the sequential decoders presented in the literature, the hybrid SCRAM decoder jointly resolves collisions, and decodes the LDPC codewords, in a similar analogy to Belief Propagation (BP), on a joint three-layer Tanner graph. The graph comprises a set of variable nodes, representing the symbols of the accommodated users, and two sets of check nodes; LDPC- and SA- check nodes. In essence, the joint SCRAM decoder induces a virtual handshake between the SA- and the LDPC- check nodes, that leads to traversing valuable information along the three-layer Tanner graph. This novel decoding structure, leads to faster convergence of the iterative decoding, along with vast superiority in the achievable PER. The results presented herein, highlight the drastic outperformance of the proposed SCRAM over the leading state-of-the-art CD-NOMA schemes. In addition, SCRAM exhibits vast superiority over the state-of-the-art Turbo-like decoding schemes (Iterative Detection and Decoding). Turbo-like decoding could be regarded as an iterative sequential scheme, that involves a feedback link between the Multiuser Detector (MUD) and the bank of FEC decoders. The SCRAM decoder, on the other hand, is strictly an iterative joint decoding scheme, such that every decoding iteration concurrently involves both the SA- and the LDPC- check nodes. Additionally, a further enhanced version of SCRAM, dubbed Collision Diversity (CoD) SCRAM is proposed. The proposed CoD SCRAM involves transmitter-side optimization, by devising enhanced channel access patterns. The CoD SCRAM follows in the footsteps of the proposed Random Access SCRAM in the sense that it still supports the highly overloaded scenarios, in addition to maintaining the joint decoding structure on the three-layer Tanner graph. Meanwhile, the proposed CoD SCRAM aims at replacing the Random Access approach adopted by SCRAM, with enhanced channel access patterns, at the price of subtle coordination at the transmitter side. The enhanced channel access patterns, adopted by CoD SCRAM, are devised in light of three main pillars. First, from an information theoretic perspective, CoD SCRAM aims at maximizing the mutual information, in order to enhance the attainable spectral efficiency. Secondly, due to the analogy between the two-layer Tanner graph of classical LDPC codes, and the three-layer Tanner graph of the joint SCRAM decoder, the well-established techniques that analyze the graphical structure of the LDPC Tanner graph, were utilized to devise the access patterns of the CoD SCRAM. Thirdly, CoD SCRAM aims at leveraging the collisions among the accommodated users, in such a way that enhances the traverse of LLRs on the joint three-layer Tanner graph. The simulation results highlight the superiority of the CoD SCRAM over the Random Access SCRAM, where the latter was already proven to outperform the leading state-of-the-art techniques presented in the literature. Moreover, the comparison of the performance of the proposed CoD SCRAM to the theoretical capacity bounds sheds the light over a fundamental finding. The proposed CoD SCRAM outperforms the Sphere-Packing Bound (SPB), calculated at the respective information block length of the underlying LDPC code. This means that the CoD SCRAM leverages the collisions among the transmitted packets, such that the joint SCRAM decoder on the three-layer Tanner graph regards the collided packets as one long codeword.