Opportunistic monitoring of pavements for noise labeling and mitigation with machine learning

Abstract

Currently, municipalities assess rolling noise on road surfaces using Close-Proximity measurements (CPX). To avoid these labor-intensive measurements, an opportunistic approach based on commodity sensors in a fleet of cars, is proposed. Blind sensor calibration eliminates the effect of measurement vehicle and varying observation conditions.Calibration relies on spatial coherence: modifiers and confounders do not interact strongly with location while the quantity of interest depends on location and less on measurement vehicle. Generalized additive speed models, car offset and de-noising autoencoders (DAE) were investigated. DAE achieves prominent results: (1) ratio of variability of measurements at a single location to the variability of measurements over all locations increases, (2) convergence of mean measurement at a location is faster, and (3) seasonal effects are eliminated. Finally, although the proposed method includes a diversity of tires, below 1600 Hz its results differ from CPX less than the difference between bi-annually repeated CPX measurements.

A scalable, self‐supervised calibration and confounder removal model for opportunistic monitoring of road degradation

“…This paper builds upon the earlier work presented in Van Hauwermeiren et al. (2019, 2021). The work presented here extends the proposed methodology to a scalable approach that allows to include a large number of measurement vehicles into the proposed road monitoring system.…”

“…Opportunistic monitoring of road surfaces has previously been studied at Ghent University, where a fleet of cars was equipped with sound and vibration sensor boxes. The collected data have already been used to measure the influence of pavement characteristics on rolling noise on roads (Van Hauwermeiren et al., 2021). In this work, a denoising autoencoder (DAE) (Goodfellow et al., 2016), is introduced for calibrating noise measurements made by a fleet of cars and for removing modifiers and confounders (vehicle speed, acceleration, and temperature).…”

Monitoring trends in road surface impact on rolling noise emissions

“…For reference conditions a temperature of 15°C, target speed at v85 of the road segment and zero horizontal acceleration are used. 5 Most of the observations are made between December 2020 and April 2021. Temperatures with mean 11.8 °C and standard deviation of 7.4°C occurred in the dataset.…”

“…To illustrate the effect of the denoising auto-encoder (DAE) on the noise data, the mean of the variance of different measurements within a 20 m segment, w, is compared to the variance of the mean of measurements over multiple segments, o, before and after applying the DAE (see [5] for formulas). These variables are calculated for the month March 2021 including around 1.e6 observations and results are shown in Figure 1.…”

“…Hence an appropriate on the fly calibration and denoising method is needed to support the effect of averaging and increase convergence rate. In [5] we compared several methods for calibration and denoising and concluded that a denoising autoencoder, a technique based on deep learning, seems most efficient in removing dependence on measurement vehicle and tires as well as dependence on air temperature. This pavement monitoring concept has now been deployed in several vehicles.…”