Smartphone-based Road Condition Sensing using Machine Learning techniques

To ensure the safety and facilitation of transportation, the road condition should be evaluated continually for determining the maintenance plan and reducing the risk of accidents. In this paper, we propose a low-cost and automated road condition sensing method that uses smart phone as the sensing device and exploits machine learning as the evaluator. To evaluate the road quality, the vibration, vehicle location and vehicle speed are gathered via an accelerometer of the smartphone. Then, this data is preprocessed and fed to the machine learning model for prediction the road condition. Moreover, we compare the road condition prediction performance of three machine learning models including MaxEnt, XGBoost and Support Vector machine (SVM) where the XGBoost shows the best prediction performance among others. Therefore, we implement the XGBoost for our automated road condition sensing system. Then, our proposed device can be used to evaluate the road condition with short evaluation time, high accuracy and low-cost.