PaveX, a Purdue University-related technology startup, has received a $305,000 Phase I Small Business Innovation Research grant from the U.S. National Science Foundation to develop an artificial intelligence system designed to transform how roads are inspected. The 18-month project will focus on a low-cost, vehicle-mounted platform that can collect street-level data, identify pavement damage and generate condition ratings for transportation agencies. By replacing slow, labor-intensive inspection methods with automated sensing and analysis, the company hopes to make road maintenance more frequent, precise and affordable.
Road agencies traditionally rely on field crews to drive along road networks, record visible defects and determine which streets require repair. That process can be expensive and difficult to scale, particularly when agencies are responsible for thousands of miles of pavement. PaveX’s system is designed to use mobile sensors mounted on ordinary vehicles, allowing roads to be surveyed while the vehicle is already traveling through a community. Artificial intelligence then processes the collected imagery and other sensor data to identify pavement distress, including cracks, potholes and related signs of deterioration.
Mohammad Jahanshahi, PaveX’s founder and chief executive officer, said the project’s broader goal is to create an infrastructure-monitoring system that can be deployed by transportation agencies at the state, national and international levels. The platform is intended to produce practical road condition ratings that help public works departments prioritize repairs before defects become more severe. Earlier intervention could reduce long-term maintenance expenses, limit vehicle damage caused by deteriorating pavement and improve road safety for drivers, cyclists and pedestrians.
Jahanshahi is an associate professor in Purdue’s Lyles School of Civil and Construction Engineering and holds a courtesy appointment in the Elmore Family School of Electrical and Computer Engineering. He developed the underlying artificial intelligence platform and disclosed the technology to Purdue Innovates’ Office of Technology Commercialization. Purdue applied for patent protection for the intellectual property and granted PaveX an exclusive license to commercialize it, linking university research in civil engineering, computer vision and transportation systems with a startup intended to bring the technology into widespread use.
A central focus of the NSF-funded project will be PaveX’s automation framework, which is designed to provide comprehensive, lane-level coverage while reducing two persistent bottlenecks in road assessment. The first is route planning: agencies must determine how inspection vehicles can cover complex networks efficiently without overlooking important streets. The second is the manual validation of uncertain detections, a process in which human reviewers examine images flagged by an algorithm. By automating more of both tasks, PaveX aims to increase the amount of roadway that can be evaluated while reducing the time required from engineering staff.
The project will also investigate a Bayesian multiframe data-fusion method to improve the reliability of pavement-distress detection. In a conventional image-analysis system, each frame captured by a moving vehicle may be evaluated independently. That approach can create false positives when shadows, stains, road markings or temporary objects resemble pavement damage. PaveX’s proposed method will combine observations from consecutive frames, using repeated evidence across time to determine whether a suspected defect remains consistent as the vehicle moves past it. In statistical terms, the Bayesian framework can update the probability that a feature represents genuine pavement distress as additional observations become available.
This temporal fusion could help the system distinguish persistent structural defects from visual noise. A crack that appears in a similar position across multiple frames is more likely to be a real road feature than an artifact visible in only one image. Conversely, inconsistent signals can be down-weighted, reducing false alarms and the need for human review. The approach is particularly important in real-world environments, where lighting, weather, vehicle speed, camera motion, traffic and pavement textures can vary substantially from one survey to the next.
PaveX will measure the success of the first phase through reductions in route-planning overhead and improved detection performance under these real-world conditions. Since January 2025, the company has surveyed more than 6,200 miles of roads in Indiana, Michigan, North Carolina, Utah and California. Those deployments have provided experience across different climates, pavement types and transportation environments, giving the team an expanding body of roadway data with which to test and refine its algorithms.
The NSF award represents a step toward scaling the technology across larger transportation networks, but the company’s ambitions extend beyond dedicated inspection vehicles. Jahanshahi said PaveX could eventually be integrated into autonomous vehicles, allowing cars and other mobile platforms to collect road-condition data continuously as they travel. Such a distributed sensing network could give infrastructure agencies more frequent updates than periodic inspections, revealing how pavement changes over time and helping officials direct limited repair budgets to the locations where they are most urgently needed. If the system can demonstrate reliable performance at scale, routine travel itself could become a source of continuously updated infrastructure intelligence.
Subject of Research: Artificial intelligence, road infrastructure assessment, pavement-distress detection, mobile sensing and transportation engineering
Article Title: AI-Powered Road Survey System Wins $305,000 NSF Grant to Automate Pavement Inspections
Web References: PaveX; Mohammad Jahanshahi at Purdue University; Purdue Innovates Office of Technology Commercialization
References: National Science Foundation Small Business Innovation Research grant; Purdue University; PaveX
Image Credits: Purdue University photo/Kevin Crisp
Keywords: Artificial intelligence, road infrastructure, pavement monitoring, transportation engineering, roads, streets, road construction, computer vision, smart cities, autonomous vehicles, National Science Foundation, PaveX
