F$^4$-CKM: Learning Channel Knowledge Map with Radio Frequency Radiance Field Rendering

TL;DR

The work tackles CSI acquisition bottlenecks in 6G by proposing F$^4$-CKM, a CKM construction framework that renders an RF radiance field of the environment and predicts downlink CSI from uplink CSI. It introduces the WiRARE network with a spatial-aware backbone and frequency-aware FCA modules, a shaping filter for data augmentation, and a spherical Fibonacci Grid-based angular sampling to enable location-free, fast-learning CKMs. Empirical results on simulated indoor datasets and the Argos real-world dataset show substantial gains in PSNR, SGCS, and spectral efficiency compared with NeRF-based and model-based baselines, with strong robustness to uplink CSI imperfections. The approach offers practical benefits for 6G deployments by reducing CSI feedback overhead, improving beamforming fidelity, and enabling scalable CKM construction, while highlighting directions for outdoor dynamics and online recalibration as future work.

Abstract

In 6G mobile communications, acquiring accurate and timely channel state information (CSI) becomes increasingly challenging due to the growing antenna array size and bandwidth. To alleviate the CSI feedback burden, the channel knowledge map (CKM) has emerged as a promising approach by leveraging environment-aware techniques to predict CSI based solely on user locations. However, how to effectively construct a CKM remains an open issue. In this paper, we propose F$^4$-CKM, a novel CKM construction framework characterized by four distinctive features: radiance Field rendering, spatial-Frequency-awareness, location-Free usage, and Fast learning. Central to our design is the adaptation of radiance field rendering techniques from computer vision to the radio frequency (RF) domain, enabled by a novel Wireless Radiator Representation (WiRARE) network that captures the spatial-frequency characteristics of wireless channels. Additionally, a novel shaping filter module and an angular sampling strategy are introduced to facilitate CKM construction. Extensive experiments demonstrate that F$^4$-CKM significantly outperforms existing baselines in terms of wireless channel prediction accuracy and efficiency.

Figures (17)

• Figure 1: Propagation modeling for the radiated signal from a wireless radiator.
• Figure 2: Decomposition of the wireless radiator response.
• Figure 3: Variations in the propagation distances of signals arriving at different receive antennas.
• Figure 4: The construction pipeline of F$^4$-CKM. The uplink CSI at the UE's antenna array is taken as input. We first perform angular sampling, followed by radial sampling along each ray. Subsequently, replicas of the original uplink CSI are fed into the shaping filter module, along with two flows of side information that serve as guidance. This module processes the CSI replicas into augmented queries for the WiRARE network. WiRARE then leverages these queries to obtain the aggregating coefficients and material properties of the sampled radiators, which are aggregated to form the final prediction for the downlink CSI.
• Figure 5: The SFG sampling algorithm. (a) illustrates the mechanisms of SFG sampling. (b) provides top views of $648$ sampled rays using LL sampling and SFG sampling, respectively. The angular resolution for LL sampling is set to $10^{\circ}$.