Pavement management relies on different types of data. Inventory data describe the physical elements of a road system. Pavement construction history data describe the pavement structure and layer composition. Condition data describe the condition of elements that can be expected to change over time. Traffic and climate data describe the traffic and environmental loading the pavement is subjected over time. Cost data describe historical investments and serve as a data source for future cost estimates.
There are a wide range of technologies available to the highway engineers for measuring attributes of the pavement network and they are advancing at a rapid pace. Many of these technologies operate at traffic speed collecting massive amount of data effectively with little or no impact to highway users. There is increased availability of auxiliary data sources such as Modern-Era Retrospective analysis for Research and Applications (MERRA) and MERRA-2 climate data from NASA’s Global Modelling and Assimilation Office, USDA’s Natural Resources Conservation Service (NRCS) soil survey database to estimate subgrade soil engineering properties, data from connected and automated vehicles, and data that are crowdsourced. However, pavement management applications have not been able fully utilize this rapidly expanding amount and type of “big data.” Traditional analysis procedures used in pavement management are not readily adaptable to the 5 V’s of big data - volume, velocity, variety, veracity and value. This is not unique to pavement management. Recent advances in the field of Data Science, availability of open source tools implementing artificial intelligence (AI) techniques such as deep learning (DL) (for example, Python DL libraries TensorFlow and Caffe) and advances in data storage and processing power provides an opportunity to complement proven traditional analysis approaches with data science and DL techniques to better leverage big data for more informed pavement management.
The selected candidate will begin with the 1) exploration of the availability and variety of both pavement related and other publicly available data that would inform or add value to current and unexplored pavement management applications, 2) assemble/fuse identified datasets (or a subset) for data cleansing, processing, etc., 3) apply physics guided DL techniques to identify and develop causative relationships that will enhance current and innovate new pavement management applications and improve pavement performance and needs predictions, and 4) incorporate those findings into analysis methodologies and tools to transfer them to practice.
The selected candidate will collaborate with staff members throughout various relevant offices within FHWA and with those involved in Forum of European National Highway Research Laboratories (FEHRL)’s Big data for smart pavement management (BD-Pave) effort that FHWA is collaborating with. The selected candidate is expected to be an independent thinker and a self-starter. Skilled analyst with innovative ideas and talents in analyzing big datasets fused from different sources and formats, both structured and unstructured, are encouraged to apply. In addition to high level of proficiency in quantitative data analysis, the candidate should also have sufficient understanding of highway agency pavement management applications. The selected candidate should also have computer programming skills in languages such as Python.
Pavement management; Big data; Data science; Artificial intelligence; Deep learning; Data fusion; Data analytics