Model and Data Dual-Driven: A Promising Receiver Design for 6G Wireless Communications

As an emerging paradigm for 6G intelligent communications, deep learning (DL) has delivered remarkable performance gains in physical layer communications and has found its way into many applications. The existing works are categorized as model- and data-driven approaches, where the unexplainable and unpredictable nature of data-driven methods always hinder their practical deployment, and model-driven DL approaches deteriorate when the model or prior information is particularly inaccurate. Therefore, this article introduces a model and data dual-driven receiver structure for 6G communication, which aims to incorporate both wireless domain knowledge and versatile data-driven networks. We elucidate the benefits and power of such dual-driven structures and discuss recent innovations in dual-driven approaches for physical receiver design. Finally, potential research directions for dual-driven approaches are highlighted to boost future 6G intelligent physical layer communications.