Estimate Road Roughness Using Smartphone Response Data – A Semi-Supervised Learning Approach

Abstract: The smartphone-collected vehicle response data is being used to estimate the International Roughness Index (IRI). Among the existing methods, the machine learning approach is drawing attention. However, surveying the ground-truth IRI is expensive and there is limited labelled data. In contrast, there exists a wealth of unlabelled response data from road users. This scenario presents opportunities for exploring semi-supervised learning (SSL) algorithms, which have been insufficiently researched. This study addresses this gap by applying an SSL framework to refine an IRI estimation model. Our results show that the SSL-trained model achieves a lower RMSE than the fully supervised trained model.