Predictive Rate Selection for Ultra-Reliable Communication using Statistical Radio Maps

TL;DR

This paper introduces a method using statistical radio maps and Gaussian processes to predict channel statistics for ultra-reliable communication, reducing training time and improving rate selection in URLLC scenarios.

Contribution

It presents a novel approach leveraging spatial correlation and Gaussian processes to predict channel quantiles, aiding rate selection for URLLC.

Findings

Reduces training estimation phase in channel prediction

Improves rate selection accuracy for URLLC

Demonstrates effectiveness with synthetic urban micro-cell data

Abstract

This paper proposes exploiting the spatial correlation of wireless channel statistics beyond the conventional received signal strength maps by constructing statistical radio maps to predict any relevant channel statistics to assist communications. Specifically, from stored channel samples acquired by previous users in the network, we use Gaussian processes (GPs) to estimate quantiles of the channel distribution at a new position using a non-parametric model. This prior information is then used to select the transmission rate for some target level of reliability. The approach is tested with synthetic data, simulated from urban micro-cell environments, highlighting how the proposed solution helps to reduce the training estimation phase, which is especially attractive for the tight latency constraints inherent to ultra-reliable low-latency (URLLC) deployments.