Artificial intelligence techniques for pavement performance prediction: a systematic review

Abstract

Pavement performance prediction is a major part of a pavement management system that directly impacts the effectiveness of maintenance and rehabilitation decisions. The prediction methods are commonly based on empiricism and traditional statistical models. In recent years, the application of Artificial Intelligence (AI) techniques for pavement performance prediction has gained momentum. These advanced techniques have shown promising results in civil and infrastructure analysis and asset management. With the help of AI, the accuracy and efficiency of pavement performance data analysis are able to be further improved. In this article, a systematic literature review of the existing studies related to pavement performance prediction with supervised AI and ML techniques was conducted. Articles that predicted pavement performance using image processing and computer vision methods were excluded. A total of 1370 peer-reviewed articles from IEEE Xplore, ACM Digital Library, TRID, and Scopus were initially identified, 158 of which met all inclusion and exclusion criteria and were included for the review. PRISMA guidelines were followed for conducting and reporting the review. Neural networks were the most commonly used algorithms, and the majority of the articles focused on flexible pavements and predicting the International Roughness Index (IRI), followed by Rutting. We present a summary of the algorithms, databases, input and output variables used in the previous models and discuss the existing research gaps and directions for future work.

Keywords:

• Systematic review
• pavement performance prediction
• artificial intelligence

Acknowledgments

This study was part of the ‘Towards Smart Pavements in Canada’ project under the AI4L-102. The authors would like to thank the Artificial Intelligence for Logistics (AI4Losistics) program of the National Research Council (NRC) of Canada for providing financial support for this research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by National Research Council Canada [AI4L-102].