SCMA-TPGAN: A new perspective on sparse codebook multiple access for UAV system

Abstract

As more and more devices are connected to wireless networks, channel resources are getting tighter and tighter. UAVs, as a new type of mobile device, will play a larger role in the future and will require a larger connection size. Sparse codebook multiple access as a non-orthogonal technique in the code domain can exactly solve this problem. In order to improve the bit error rate performance of sparse code multiple access (SCMA) access systems in uplink Rayleigh fading channels, we propose a new deep learning-based SCMA encoding and decoding scheme. At the SCMA encoder, the Transformer is used as the generator of the generative adversarial network model (GAN), in order to solve the problem that the sequence of the generator is too long in the information processing process. An additional noise layer is introduced at the input of the encoder, resulting in a robust representation of the encoder output, improving the noise immunity of the system. At the decoder, PatchGAN is used as the discriminator of the generative adversarial network to reduce the amount of network model parameters and computation. Simultaneous equalization network and multi-user detection network constitute the decoder network. At the same time, an attention mechanism is added between the generator and the discriminator of the generative adversarial network model to enhance the details of the information part and improve the bit error rate performance. The scheme proposed in this paper is composed of a network-assisted decoder balanced network and an autoencoder-based generative contradiction network (SCMA-TPGAN). Through experimental modeling and comparative analysis of different methods, we conclude that SCMA-TPGAN can reduce detection time and improve system bit error rate performance.