Performance Analysis of an Autoencoder-based Communication System with Numerous Transmitter-Receiver Pairs

With the significant increase in the number of users, there is a high demand to design a shared communication system that allows multiple users to communicate at the same time. In addition, the autoencoder model has recently emerged as a promising solution to automatically modulate the transmitted symbols. Therefore, in this work, we study an autoencoder-based communication system when multiple users share the same communication channel. Specifically, transmitters jointly modulate the generated symbols using the autoencoder model and send the encoded data to the receivers over the shared communication channel. The optimal constellation diagram is learned by minimizing the difference between the transmitted and reconstructed symbols. We have conducted extensive experiments to evaluate the performance of the autoencoder-based communication system sharing between multiple users. With the same data throughput, the AE-based shared communication system for multiple transmitters and receivers results outperforms traditional time-division and frequency-division multiple access control methods in terms of symbol error rate.