ASC-Net: Deep Learning for Downlink Control Information Compression

Abstract

With the rapid growth of connected devices in the sixth generation (6G) era, the overhead of control signaling faces severe pressure. Consequently, downlink control information (DCI) compression becomes an effective approach to alleviate control channel congestion while maintaining scheduling flexibility. To achieve this objective, an asymmetric sequential compression network (ASC-Net) consisting of an encoder and a decoder is proposed in this paper. At the Advanced NodeB (aNB), the encoder integrates the current DCI, slot index, and the history hidden feature to generate the compressed code. At the user equipment (UE), the decoder recovers the DCI by combining the received compressed code with the local history DCI. Additionally, the aNB deploys a local decoder identical to that at the UE, which could pre-verify the reconstruction. If there exists an error in the reconstructed DCI, a fallback scheme is employed, where the raw DCI is transmitted to ensure correctness. Experimental results show that ASC-Net achieves competitive compression efficiency while generating a fixed-length compressed code for specific deployment scenarios. Compared to variable-length coding schemes, this fixed-length property reduces blind detection complexity at the UE.