Integration of deep learning techniques and sustainability-based concepts into an urban pavement management system

Abstract

Pavement management systems (PMS) were typically developed to manage interurban networks. This paper presents an urban pavement management system (UM-PMS) that integrates all the modules of a PMS considering the particularities of the urban context. The tool uses a Geographic Information System (GIS) to import, analyze, and manage the data of the urban road network. The inspection data is obtained by an automatic equipment composed by a video camera installed on a vehicle. Images are analyzed by deep learning techniques based on Convolutional Neural Networks. Further, appropriate decisions on maintenance treatments are made by integrating multi-objective optimization and multi-criteria decision-making methods to plan efficient maintenance strategies considering economic, environmental, social, and performance objectives. Finally, the tool is applied to an urban case study to illustrate its applicability. Outcomes indicate that the proposed framework can obtain a sustainable short-term plan without losing sight of the long-term efficiency.