Machine Learning Approaches to Road Surface Anomaly Assessment Using Smartphone Sensors

Abstract

Road surface quality is an essential component of roadway infrastructure that leads to better driving standards and reduces risk of traffic accident. Traditional road condition monitoring systems fall short of current need for quick responses to maintain road quality. Several alternative systems have been proposed that utilize sensors mounted on vehicles and with the ubiquitous use of smartphone for personal use and navigation, smartphone based road condition assessment has gained prominence.

We propose to analyze different multiclass supervised machine learning techniques to effectively classify road surface conditions using accelerometer, gyroscope and GPS data collected from smartphones. Our work focusses on classification of three main class labels- smooth road, pothole and deep transverse cracks. We investigate our conjecture that using features from all three axes of the sensors provide more accurate results as compared to using features from only one axis. We also investigate the performance of deep neural networks to classify road conditions with and without explicit manual feature extraction. Our results consistently show that models trained with features from all axes of the smartphone sensors perform better than models that use only one axis. This shows that there is information in the vibration signals along all three axis for road anomalies. We also observe that the use of neural networks provide significantly accurate data classification. The approaches discussed here can be implemented on a larger scale to monitor road for defects that present a safety risk to commuters as well as provide maintenance information to relevant authorities.