Deep Learning for Cross-Technology Communication Design

Abstract: Recently, it was shown that a communication system could be represented as a deep learning (DL) autoencoder. Inspired by this idea, we target the problem of OFDM-based wireless cross-technology communication (CTC) where both in-technology and CTC transmissions take place simultaneously. We propose DeepCTC, a DL-based autoencoder approach allowing us to exploit DL for joint optimization of transmitter and receivers for both in-technology as well as CTC communication in an end-to-end manner. Different from classical CTC designs, we can easily weight in-technology against CTC communication. Moreover, CTC broadcasts can be efficiently realized even in the presence of heterogeneous CTC receivers with diverse OFDM technologies. Our numerical analysis confirms the feasibility of DeepCTC as both in-technology and CTC messages can be decoded with sufficient low block error rate.