NeWRF: A Deep Learning Framework for Wireless Radiation Field Reconstruction and Channel Prediction

TL;DR

NeWRF addresses the challenge of costly dense site surveys by predicting wireless channels at arbitrary indoor locations from sparse measurements. It extends Neural Radiance Fields to the radio frequency domain by learning a wireless radiation scene with a 5D MLP and incorporating propagation physics; it introduces DoA-guided ray casting to improve convergence and a ray-searching algorithm for inference without full DoA. The approach demonstrates that wireless radiation scenes can be learned from sparse data and provides accurate channel predictions at unvisited locations, with substantial reductions in required measurements compared to prior methods. This enables faster, cheaper, and more accurate site assessments for WiFi and 5G deployments, with potential for integration across material types and carrier frequencies.

Abstract

We present NeWRF, a deep learning framework for predicting wireless channels. Wireless channel prediction is a long-standing problem in the wireless community and is a key technology for improving the coverage of wireless network deployments. Today, a wireless deployment is evaluated by a site survey which is a cumbersome process requiring an experienced engineer to perform extensive channel measurements. To reduce the cost of site surveys, we develop NeWRF, which is based on recent advances in Neural Radiance Fields (NeRF). NeWRF trains a neural network model with a sparse set of channel measurements, and predicts the wireless channel accurately at any location in the site. We introduce a series of techniques that integrate wireless propagation properties into the NeRF framework to account for the fundamental differences between the behavior of light and wireless signals. We conduct extensive evaluations of our framework and show that our approach can accurately predict channels at unvisited locations with significantly lower measurement density than prior state-of-the-art

Figures (13)

• Figure 1: The tomography of a wireless field (channel) in a simple environment. Like water ripples, wireless signals can add up constructively or destructively, causing fine-scale spatial variation. Our algorithm, NeWRF, reconstructs this field from a sparse set of channel measurements.
• Figure 2: Wireless communication model: Wireless signals emitted by the transmitter experience attenuation, reflection, absorption, multipath interference, etc. before reaching the receiver. Our framework models all of these factors.
• Figure 3: An example environment matlab_conference and locations of the transmitter and receivers. We fix the location of the transmitter and measure the wireless channel at various receiver locations. The wireless channel is the sum of multiple propagation paths from the transmitter to the receiver.
• Figure 4: An overview of our neural channel synthesis algorithm: We backtrace each propagation path from a receiver and query an MLP with the 5D coordinates (location and viewing direction) of points along the path, to predict the radiated signal (amplitude and phase) at that point. Then we integrate signals from all paths to synthesize a channel. We model reflections as virtual transmitters at the point of reflection.
• Figure 5: A Learnt Wireless Scene: Spots with high volume density values indicate the locations of (virtual) transmitters. The gray cube shows the 3D environment model presented in Figure \ref{['fig:conference_room']}.