Deep Learning-Based Codebook Optimization for Scma in 5G and Beyond Wireless Networks

We know that in traditional Orthogonal Multiple Access (OMA) techniques, resources like frequency, time, or code are assigned to each and every user. Whereas Sparse Code Multiple Access Scheme is a Non-Orthogonal Multiple Access Scheme (NOMA) which allows multiple users to share the same resources at the same time by mapping their data onto sparse multi-dimensional codewords. This Non-orthogonal design helps in improving the spectral efficiency, system capacity and also supports massive connectivity needed for 5 G and future wireless networks. While the performance heavily depends on the quality of the hand-crafted codebooks. In this paper, we observe how an SCMA system performs with both hand-crafted codebooks and codebooks optimized through a deep learning framework. We use the Message Passing Algorithm (MPA) at the receiver side for detection under Rayleigh fading and Additive White Gaussian Noise (AWGN) channels. The system model is designed for a setup with 6 users and 4 resources. To measure the Symbol Error Rate (SER) and Bit Error Rate (BER) over different signal-tonoise ratio (SNR) levels we run detailed simulations. From the results we come to know that SCMA with optimized codebooks greatly reduces errors, helping the system to maintain reliable communication even in noisy and fading environments.