Noncoherent Reconfigurable Intelligent Surface With Differential Modulation and Reflection Pattern Training

In this letter, we propose a novel noncoherent reconfigurable intelligent surface (RIS)-assisted downlink architecture aided by lightweight reflection pattern training as well as noncohererent detection. More specifically, the proposed scheme amalgamates the benefits of reflection pattern training and differential modulation, which operate without any channel estimation in each stage of RIS optimization and data detection, hence combating the delay-induced limitations. The proposed scheme achieves low training complexity and robustness over the channel variation in comparison to the conventional RIS optimization assuming the use of channel state information. We demonstrate that the proposed scheme outperforms the conventional coherent benchmarks.