This paper addresses the problem of adaptive codebook (CB) selection for downlink (DL) precoder quantization in channel state information (CSI) reporting. The accuracy of precoder quantization depends on propagation conditions, requiring independent parameter adaptation for each user equipment (UE). To enable optimal CB selection, this paper proposes UE-assisted CB selection at the base station (BS) using reported by the UE statistical channel properties across time, frequency, and spatial domains. The reported assistance information serves as input to a neural network (NN), which predicts the quantization accuracy of various CB types for each served user. The predicted accuracy is then used to select the optimal CB while considering the associated CSI reporting overhead and precoding performance. System-level simulations demonstrate that the proposed approach reduces total CSI overhead while maintaining the target system throughput performance.
Massive Multiple-Input Multiple-Output (M-MIMO) antenna systems have become an integral part of commercial 5-th Generation New Radio (5G NR) deployments in the C-band [1]. In practice, this technology has demonstrated significant advantages, including higher spatial multiplexing gains and improved transmission directivity through the use of advanced beamforming schemes. M-MIMO, along with its evolution for upper mid-bands, is also considered a key solution for meeting the growing capacity demands of future 6-th Generation (6G) cellular networks [2].
The performance benefits of M-MIMO systems are achieved through advanced precoding schemes implemented at the base station (BS) transmitter. Two types of channel state information (CSI) are commonly considered to assist the beamforming operations. The first type relies on channel reciprocity and sounding reference signals (SRS) transmitted by the user equipment (UE). However, its efficiency is limited by uplink (UL) transmission power constraints, making it more suitable for cell-center users. The second, more widely applicable approach is codebook (CB)-based downlink (DL) precoding. In this method, the UE measures the channel response from the BS digital antenna ports using channel state reference signals (CSI-RS). The measured channel is then quantized using a predefined CB and reported to the BS via the UL control channel, enabling DL beamforming.
To support CSI compression, 5G NR specifies various types of CBs that quantize CSI across the spatial, frequency, and Doppler/time domains [3]. The level of compression and resulting UL control overhead are primarily controlled by a set of parameters defining the number of Discrete Fourier Transform (DFT) basis vectors used to represent the quantized CSI. In addition to CB-based methods, 6G system is expected to adopt AI/ML-based approaches for CSI compression, particularly using auto-encoders. This technique offers higher CSI compression levels while maintaining the similar or lower CSI quantization error. Furthermore, different levels of compression can be supported by adapting the number of auto-encoder outputs [4].
A key practical challenge in supporting various CSI quantization is the proper selection of CSI compression level and configuration of the CB parameters to ensure target CSI quantization error and UL control channel overhead. From a network perspective, the quality of CSI quantization is typically unknown at the BS. As a result, practical BS configurations should rely on a common set of CB parameters for all UEs, which may not always provide optimal performance under diverse channel conditions.
The problem of UE-specific CB selection has been discussed in several papers. More specifically, [5] proposes a basic approach based on Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) channel classification serving as foundation for coarse switching between basic CB types. In [6], a federated reservoir computing framework (CA-FedRC) for 5G NR CB adaptation is introduced, balancing performance and feedback overhead based on some CSI indicators. Using simple link-level channel models, it was shown that dynamic switching of CB under diverse communication conditions can significantly improve throughput performance while reducing UL control channel overhead. However, the corresponding method requires large number of input parameters and some convergence time to make optimal CB selection decision possibly resulting into performance degradation during transient time. In [7] an UEassistance information-based approach is proposed, leveraging time-variability characteristics of the channel between BS and UE for system parameter selection including basic switching of CBs. The proposed UE assistance information, which corresponds to the channel correlation in time domain, enables 5G network to UE-specifically select feedback parameters (i.e., type of CSI feedback) to maximize both user and system performance. However, the approach proposed in [8] relies solely on traditional single threshold-based CB selection, making it challenging to determine accurate threshold for optimal adaptation. Moreover, timedomain channel properties (TDCP) alone may not be sufficient to determine the optimal CB configurations of all parameters. A more accurate approach should consider other channel characteristics (e.g., in spatial and frequency domains) and employ joint selection method for CB adaptation. In particular, it is well known that CSI compression based on Type-1 CB in 5G NR may experience severe performance degradation in rich scattering environments (i.e., channels with high delay and angular spreads) [8]. However, in high-mobility scenario with lower channel correlation in time-domain, Type-1 CB provides more robust performance and may outperform other CSI quantization methods depending on Doppler spread. This necessitates joint consideration of all channel properties to decide on the optimal CB. Taking these considerations in
This content is AI-processed based on open access ArXiv data.