On-going Research
Improve Data Quality for Automated Pavement Distress Data Collection
Sponsor:
Project Summary: The research team will assess current automated and semi-automated data collection practices and QC/QA practices in TxDOT. Researchers will evaluate the accuracy, precision, and reliability of the pavement distress data collection methods used in the network level for TxDOT’s PMIS data collection. The team will then develop an optimized sampling method for auditing the automated pavement condition data to reliably and efficiently locate the pavements with potential data quality problems, develop sound QA statistics based on both individual ratings/indexes and Scores and provide recommendations on the threshold values to detect potential data quality issue during data collection season. Ultimately, TSUSM researchers will establish QA guidelines, procedures, or specifications for automated and semi-automated pavement condition surveys that could be used by TxDOT to improve data quality management practices for contracting pavement condition data collection.
More information:
Using artificial intelligence to improve the accuracy of automated pavement condition data collection
Sponsor: National Science Foundation
Amount Funded: $250,000
Project Summary: This project proposes a cost-effective automated pavement condition survey system with accurate and robust measurement abilities using machine learning algorithms and data enabled analytics. Pavement management engineers need to conduct pavement surface condition data collections (or surveys as termed in industry) to determine the service conditions of a highway pavement for the purposes of routine monitoring or planned corrective actions. In the past two decades, more and more state highway agencies have adopted image and laser sensor based automated or semi-automated technologies for pavement condition data collections due to the advantages of saving in labor, time, and cost. According to a preliminary market study, the mainstream technologies are still based on traditional image processing methods. While the sensors adopted in the automated surveys are quite sophisticated and expensive, the supporting rules-based image processing algorithms lack ingenious and delicate abilities. Therefore, the mainstream technologies suffer from inconsistency and discrepancy in data quality. Furthermore, data errors could be introduced from both the data collection and data analysis processes. Additional human inputs for quality assurance are needed to improve the data quality during the data management process. Meanwhile, the existing high costs of the automated technologies for pavement condition data collection are of major concerns for state and local agencies. Therefore, this proposed new system is based on improved automatic three-dimensional (3D) image data collection and processing technologies.
Artificial Intelligence for Pavement Condition Assessment from 2D/3D Surface Images
Sponsor:
Project Abstract: TxDOT started to use automated and semi-automated methods to collect pavement condition data since FY 2017. However, there are still accuracy and precision issues associated with the reliability of the existing automated and semiautomated data collection methods. Therefore, the main objective of this research is to develop data quality assurance guidelines for TxDOT to improve the quality of automated pavement condition data. The three components in the research include the development of an audit sampling method, a set of consistency check criteria for pre-analysis of new data, and data quality criteria in acceptance of new data. The cost-efficient audit sampling method and the data-enabled consistency thresholds will assist TxDOT in more effectively locating pavement sections with potential data quality issues. The quality acceptance criteria will provide TxDOT with systemic decision support in data quality. In the project, a pilot study for a selected TxDOT District will be implemented with the developed guidelines to evaluate the effectiveness of the proposed data quality assurance procedures for data quality improvement. The proposed project will assist TxDOT to enhance the accuracy, precision, and reliability of the automated pavement condition data which would eventually help the State of Texas improve its pavement performance condition.
Sponsor: Texas Department of Transportation (TxDOT)
Amount Funded: $496,746
Project Abstract: Hydroplaning occurs when a layer of water builds up between a vehicle’s tires and the pavement surface, leading to a loss of traction that prevents the vehicle from responding to control inputs such as steering, braking, or accelerating. This loss of control in the vehicle can cause severe or even fatal crashes to occur. According to data from the Road Weather Management Program for the years 2007 to 2016, there were an estimated 860,000 weather-related motor vehicle accidents, with wet pavement identified as the primary cause. These incidents led to around 320,000 injuries and 4,000 fatalities. Researchers have developed several approaches to evaluate hydroplaning potential (HP) and predict areas susceptible to hydroplaning under various environmental conditions and vehicle speeds. This was achieved by integrating pavement condition data with GIS and LiDAR information. However, the Texas Department of Transportation (TxDOT) does not collect LiDAR measurements in their annual network-level survey. Thus, there is a need to use data readily available in the pavement management systems, such as profile and texture, and combine that information with rut depth measurements and road geometrics (i.e., cross-slope) to assess the HP of roads at higher level of accuracy and reliability.
Development of deep learning-based automated data collection technology for coastal highway pavements
Sponsor: US Department of Transportation (USDOT) through CREATE UTC
Amount Funded: $115.136
Project Abstract: The harsh environmental conditions of coastal areas, including extreme weather events, saltwater corrosion, tides, winds, and waves, make maintaining these roads a significant challenge. One solution to improve the resilience of coastal roads is through improved pavement maintenance, which involves maintaining roads at the right time to extend their service life cost-effectively. However, accurately monitoring the conditions of coastal roads is a challenge. Therefore, the research problem is to identify and evaluate available technologies for pavement condition evaluation in coastal areas, and assess their efficiency, accuracy, and cost-effectiveness. The research aims to provide a solution to the challenges faced in evaluating the performance of coastal roads which can support a maintenance program specific to coastal roads. The research will improve the resilience of coastal roads, reduce repair costs, and promote the economic growth of coastal areas.
Analyzing Pre- and Post-Coastal Hazard Pavement Conditions to Optimize Response Strategies for Coastal Infrastructure Resilience
Sponsor: US Department of Transportation (USDOT) through CREATE UTC
Amount Funded: $110,988
Project Abstract: Texas’ coastal region stretches over 367 miles along the Gulf of Mexico which is a significant ecological and economic zone encompassing beaches, marshes, estuaries, and barrier islands. This area supports a vibrant tourism industry, international trade, commercial fishing, and energy production, with major ports such as Houston, Corpus Christi, and Galveston playing vital roles. However, Texas’ coastline faces increasing risks from natural hazards, necessitating efficient and effective infrastructure response strategies to mitigate impacts and ensure rapid recovery. This research aims to investigate the effects of coastal hazards on pavement conditions and to use network analysis for optimizing pavement infrastructure response, maintenance decisions, and treatment allocation to achieve equitable and resilient coastal communities. The study focuses on Houston which is a key urban center exposed to frequent coastal hazards. Hurricane Harvey was selected as a case study for in-depth analysis.