AI Networking Cookbook: Practical recipes for AI-assisted network automation and development
TL;DR
An autoencoder-based solution to compress CSI at the UE side and reconstruct it at BS and outperforms the conventional CSI feedback scheme by approximately 21% in terms of throughput is proposed.
Massive multiple-input multiple-output (MIMO) systems require accurate channel state information (CSI) at the base station, whereas frequency-division-duplex downlink CSI feedback introduces substantial uplink overhead. Autoencoder-based CSI feedback networks reduce feedback dimensionality, but their parameter size and reconstruction cost may still be non-negligible for lightweight 6G terminals. This paper proposes LightCSI-Net, a compact CSI feedback network that integrates depthwise separable convolution, squeeze-and-excitation channel attention, latent compression, and residual reconstruction. Experiments are conducted on indoor and outdoor COST2100-format small-scale channel data under compression ratios of 1/4, 1/8, 1/16, and 1/32. The evaluation uses normalized mean square error (NMSE), cosine similarity, trainable parameters, and single-sample inference time. At the 1/16 compression ratio, LightCSI-Net reduces trainable parameters by approximately 49.3% compared with CsiNet while maintaining comparable indoor NMSE and cosine similarity. However, its measured CPU inference time increases from 0.091 ms to 0.283 ms, showing that parameter reduction does not automatically imply lower latency. We therefore interpret the proposed model as parameter-and memory-efficient, and provide a hardware-aware latency discussion for URLLC-oriented deployment.
In frequency division duplex (FDD) mode, channel state information (CSI) feedback is crucial for ensuring communication quality in massive multiple-input multiple-output (MIMO) systems. To reduce the feedback overhead, the user equipment (UE) typically compresses the CSI matrix before feeding it back to the base station (BS). Notably, deep learning has been proven well-suited for the CSI feedback task, wherein auxiliary prior information can further enhance CSI compression and reconstruction. However, existing methods ignore such prior information, resulting in suboptimal performance. In the meantime, the channel knowledge map (CKM) has emerged as a key enabler that can provide multi-modal information based on UEs’ positions. In this letter, we propose a CKM-aided CSI feedback scheme. We first introduce a framework for extracting coarse angle and delay information based on the UE’s position and the environmental contour. Then, the angle and delay information are utilized through a plug-and-play (PnP) module to help the decoder focus on the significant region, which is compatible with existing DL-based CSI feedback schemes. Simulation results show that the CKM-PnP module can significantly enhance the decoder’s ability to reconstruct the CSI matrix, especially in high-compression scenarios. For reproducibility, the code is accessible at https://github.com/github-whh/Ckm-CF
A unified comparative analysis of three widely used spatially correlated centralized scattering channel models in a multi-cell massive MIMO system suggests that M-MMSE is preferable in strongly correlated environments, while lower-complexity schemes such as RZF provide a favorable trade-off in scenarios with milder spatial correlation.
Stable downlink precoding in time-division duplex (TDD) multi-user multi-input multi-output (MU-MIMO) systems critically depends on the reliability of channel state information (CSI) at the transmitter. In mobile environments, Doppler-induced channel variation, estimation noise, and pilot contamination distort the channel structure, often leading to ill-conditioned matrices and unstable linear precoding. This paper proposes a Doppler-aware CSI reliability framework in which the condition number of sounding reference signal (SRS)-based channel estimates is used as a per-snapshot indicator of spatial robustness. Instead of relying solely on instantaneous CSI, the proposed approach selects the better-conditioned CSI between current and previously estimated channel realizations prior to downlink transmission, thereby improving robustness under mobility. System-level simulations compliant with 5G New Radio specifications are conducted under fixed, increasing, and decreasing Doppler scenarios using clustered delay line (CDL) channel models. Results show consistent gains in throughput and block error rate (BLER), approaching the performance of the perfect channel state information at transmitter (CSIT). Among the considered schemes, regularized zero forcing (RZF) precoding exhibits higher robustness to CSI degradation, while block diagonalization (BD) is more sensitive to Doppler variations. Overall, the proposed method provides a lightweight and standard-compatible solution for improving MU-MIMO precoding reliability in time-varying channels.
Massive Multiple Input Multiple Output or massive-MIMO systems are a cornerstone technology in modern wireless communications, especially in 5G and beyond networks. Including Orthogonal Time Frequency Space modulation (OTFS) in Massive MIMO makes the systems more robust, with further promise for robust performance in both high-mobility and multipath-rich environments. OTFS taps into the delay-Doppler domain representation of the wireless channel, ensuring greater immunity to time-frequency variations and enhanced channel estimation compared to traditional modulation methods. While going through the detailed literature available for OTFS based Massive MIMO system it was seen that a very less work has been carried out in exploring the potential and performance enhancements of the system if the system were implemented using some diverse signal transforms such as DWT, FrFT etc, moreover it has not yet been explored as to see the effect of fading onto these systems by deployment of these diverse transforms, Therefore it leaves a gap in the available literature. This paper focusses on the improvement of the Massive MIMO-based OTFS systems by implementing the systems using more advanced signal transforms such as Discrete Wavelet Transform (DWT) and Fractional Fourier Transform (FrFT) under various fading channels. The simulation results have shown significant improvement in performance when implemented using these advanced transforms.
Conventional link adaptation typically relies on scalar link-quality indicators such as signal-to-noise ratio (SNR), while richer channel state information (CSI) can improve adaptation at the cost of higher processing complexity. This paper investigates a compact alternative for modulation selection in a real-time multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) system using channel frequency response (CFR) magnitude descriptors. A dataset of 87,817 over-the-air (OTA) samples is collected using a USRP-based testbed, with CFR measurements extracted at the base station (BS) from received uplink pilots. Decision tree (DT), random forest (RF), and k-nearest neighbours (KNN) classifiers are evaluated using BS-side SNR, CFR features, and their combination. SNR-only classifiers achieve 35%-42% test accuracy, whereas CFR-only features achieve 73.6%, 81.4%, and 80.0% for DT, RF, and KNN, respectively. CFR-based performance is maintained near the 10% BLER reliability thresholds, with RF reaching 82.8%. A depth-7 DT with 123 leaves is further integrated into the LabVIEW C Node for real-time inference. The results show that compact BS-side CFR descriptors provide more discriminative information than the available scalar BS-side SNR while remaining suitable for lightweight SDR implementation.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.