Knowledge Base-Based High Compression Ratio CSI Feedback for RIS-Assisted mmWave Communications

Reconfigurable intelligent surface (RIS) has emerged as a pivotal technology in future wireless communication, which can expand millimeter wave signal coverage and enhance communication system performance. In frequency division duplex systems, the user equipment is required to provide channel state information (CSI) to the base station through a feedback channel. However, a substantial number of unit cells in RIS results in increased CSI feedback, consuming bandwidth and time resources. This correspondence proposes a knowledge base-based high compression ratio CSI feedback scheme (Knowledge Base Network, KBNet) for millimeter wave RIS-assisted wireless communication. The scheme employs a learnable knowledge base to approximate the distribution of CSI features and only needs to transmit the indexes of the vectors in the knowledge base that are most similar to the CSI feature, which can achieve accurate recovery of CSI under the condition of improving compression ratio. Simulation results demonstrate that the proposed scheme significantly improves channel recovery accuracy compared to benchmark methods under high compression ratios while efficiently reducing model size and computational complexity.