Evolution of prediction models for road surface irregularity: Trends, methods and future

Abstract

In the modern era, the importance of prioritizing traffic safety has become increasingly evident, requiring dedicated focus. An effective strategy for improving traffic safety involves optimizing road roughness to minimize road bumps and mitigate the risk of accidents. Currently, artificial intelligence algorithms are widely recognized for their capacity to accurately forecast pavement roughness in intricate environments. The current state of research on road roughness prediction using artificial intelligence approaches is found to be deficient in providing a comprehensive review. This paper aims to provide a comprehensive analysis of the patterns in predicting pavement roughness using artificial intelligence algorithms through a systematic review. This article provides an overview of the development process of IRI prediction and introduces commonly used artificial intelligence methods in the road field. These methods are primarily categorized into machine learning and deep learning. The article also presents a comprehensive overview of the similarities and differences among various works in this domain. Regarding the issue of data sources, it is divided into LTPP database and other databases, summarizing the data sources and volume used in the literature, as well as independent variables including road age, material property, road performance, climate parameters, etc. The challenges and future perspective in predicting road International Roughness Index (IRI) for the future are proposed, taking into consideration the complexity of data collection and limitations on the development of artificial intelligence networks.