Core Concepts
ML-enabled RIS configuration with domain knowledge for improved performance and scalability.
Abstract
This article introduces RISnet, a neural network architecture for optimizing reconfigurable intelligent surfaces (RIS) with mutual coupling and partial CSI. It addresses challenges like scalability, channel estimation, and joint optimization of precoding and RIS configuration. The work combines machine learning and domain knowledge for improved wireless communication performance.
Introduction to RIS optimization with ML and domain knowledge.
Challenges of mutual coupling, scalability, and partial CSI.
Proposal of RISnet architecture for joint optimization.
Application of ML for RIS configuration and precoding.
Comparison with existing analytical methods.
Stats
"More than 1000 RIS elements" (line 41)
"16 out of 1296 elements" (line 53)
"36x36 RIS elements" (line 191)
Quotes
"An early contribution to combine ML technique and domain knowledge in communication for NN architecture design." (line 41)
"RISnet can configure the phase shifts of all RIS elements with the partial CSI of a few RIS elements if the channel is sparse." (line 67)