Multi-Node Vehicular Wireless Channels: Measurements, Large Vehicle Modelling, and Hardware-in-the-Loop Evaluation

TL;DR

This paper introduces a novel mobile multi-node vehicular wireless channel sounding system, analyzes large vehicle scenarios, and validates a stochastic channel model through measurements and hardware-in-the-loop testing, aiding future safety applications.

Contribution

It presents the first fully mobile multi-node vehicular channel sounding system and develops a geometry-based stochastic model validated against real measurements.

Findings

Path loss, RMS delay spread, and Doppler spread deviations are within 3.6 dB, 78 ns, and 52 Hz for 80% of the time.

Packet error rate from measurements aligns with model predictions for 86% of the duration.

The model accurately captures large vehicle effects in dynamic vehicular environments.

Abstract

Understanding multi-node vehicular wireless communication channels is crucial for future time-sensitive safety applications for human-piloted as well as partly autonomous vehicles on roads, railways and in the air. These highly dynamic wireless communication channels are characterized by rapidly changing channel statistics. In this paper we present the first fully mobile multi-node vehicular wireless channel sounding system, which is capable of simultaneously capturing multiple channel frequency responses, ensuring that measurement conditions are identical for all observed links. We use it to analyze road scenarios with multiple vehicles and a large obstructing double-decker bus. The empirical measurement data is used to parametrize a model for the large obstructing vehicle within a geometry-based stochastic channel model. We compare the time-variant channel statistics obtained from our channel model with the ones from the measurement campaign. By means of a channel emulator we obtain the packet error rates of commercial modems for the measured and the simulated wireless communication channels and compare them, in order to validate the model at the link level. We find that the path loss, the root mean square (RMS) delay spread, and the RMS Doppler spread deviate by less than $3.6\,\mathrm{dB}$, $78\,\mathrm{ns}$, and $52\,\mathrm{Hz}$, respectively, for $80\%$ of the total simulation duration. The PER obtained from measured data is within the maximum and minimum bounds of our model for $86\%$ of the simulation duration.