Featured Projects

ATIC conducts applied research to advance the inspection, assessment and construction of transportation infrastructure. Its projects combine cutting-edge technology with real-world field data to develop practical tools and datasets for engineers and agencies.

Bridge Inspection and Monitoring

Bridge Inspection and Monitoring is a project that uses unmanned aerial systems equipped with high-resolution cameras and infrared sensors to document bridge conditions and identify signs of delamination and deteriorating concrete. UAS-collected imagery is processed into orthomosaics, 3D point clouds, and thermal overlays that give engineers a detailed, non-contact view of bridge deck health without requiring lane closures or placing inspectors in hazardous conditions.

Key highlights include:

• Captures both surface-level and subsurface anomalies using visual and infrared sensors.
• Produces georeferenced orthomosaics, 3D models and thermal imagery for detailed condition documentation.
• Reduces safety risks for inspection personnel by eliminating the need for close physical access.
• Sponsored by NDDOT, CTIPS and Collins Engineers, Inc.

AI-Driven Crack Detection

The AI-Driven Crack Detection project develops an end-to-end automated workflow for detecting and mapping cracks on bridge decks using drone imagery and deep learning. High-resolution UAS images are annotated at the pixel level and used to train a semantic segmentation model that identifies crack locations, classifies crack types and documents severity. A lightweight web-based interface allows inspectors to visualize and interact with model outputs in a standardized, repeatable format.

Key highlights include:

• Automates crack detection at the pixel level, replacing time-consuming and subjective manual inspection.
• Produces crack probability maps and binary masks suitable for measurement and reporting.
• Uses multiple annotators per image to improve model robustness and generalization.
• Supports standardized documentation and data-driven maintenance decision making.

Winter Roadway Surface Condition Sensing

The Winter Roadway Surface Condition Sensing project proposes a non-contact sensing system to classify hazardous winter road conditions, including ice, snow, slush and residual brine, based on how they reflect light. Using optical and hyperspectral sensing techniques, the system evaluates surface reflectance to improve real-time condition awareness for winter maintenance operations. A particular focus is detecting black ice and quantifying how long anti-icing brine treatments remain effective as conditions change.

Key highlights include:

• Addresses a critical gap: current tools cannot reliably detect thin clear ice or measure brine treatment effectiveness over time.
• Pairs laboratory testing with field validation across varied pavement types and surface conditions.
• Develops predictive models linking reflectance signals to surface condition and treatment performance.
• Sponsored by CTIPS, DTN, NDDOT and UND.

Railroad Ballast Moisture Monitoring

The Railroad Ballast Moisture Monitoring project investigates hyperspectral imaging as a non-contact method to detect and quantify moisture trapped in railroad ballast, the crushed stone foundation that supports railroad tracks. Excessive moisture weakens track support, accelerates fouling, and increases derailment risk, yet current inspection methods rely on visual assessment and cannot evaluate subsurface drainage conditions. The project is building a spectral library for ballast materials and developing both physics-based and AI-driven models to predict moisture conditions from reflectance data.

Key highlights include:

• Uses hyperspectral imaging to reveal moisture and degradation invisible to the human eye.
• Validated using ballast materials collected from a real-world derailment site.
• Includes an outdoor seven-tie laboratory track section and a rail-mounted cart for controlled field testing.
• Sponsored by the Federal Railroad Administration.

Additive Construction of Culverts

The Additive Construction of Culverts project evaluates concrete 3D printing as a solution to the growing culvert replacement backlog across North Dakota, where aging infrastructure and workforce shortages push replacement timelines to 24 to 36 months. The project develops printable concrete mix designs optimized for structural performance and has successfully fabricated and load-tested full-scale printed pipe culverts and box culvert formwork. Complementary finite element modeling work validates structural behavior and guides future design.

Key highlights include:

• Addresses workforce shortages and supply chain constraints in culvert construction through automation.
• Successfully printed and structurally tested a 33-inch diameter pipe and a 3x4-foot box culvert formwork.
• Combines experimental construction with advanced numerical modeling in Abaqus for comprehensive validation.
• Sponsored by NDDOT, with ATIC among the world’s pioneers in 3D-printed concrete culverts.

Smart UAS for Ancillary Structures Condition Assessment

The Smart UAS for Ancillary Structures Condition Assessment project is a research initiative focused on using small unmanned aerial systems (drones) to inspect highway ancillary structures, things like signs, lighting poles, signal masts and overhead bridges, more quickly and affordably than traditional methods.

The project’s core innovation is a custom-designed UAS payload capable of real-time, autonomous defect detection using deep learning AI models. The system can identify three key defect types on the fly: corrosion, missing bolts and cracks, without requiring manual review during flight. A cyber-physical interface was also developed to keep human inspectors in the loop, giving them control over and visibility into the AI’s detection decisions.

Key highlights include:

• Addresses a major gap in existing UAS technology, which lacks real-time AI-driven defect detection.
• Enables faster, lower-cost inspections without requiring lane closures or scaffolding.
• Combines deep learning with drone hardware into a purpose-built inspection payload.
• Balances automation with human oversight through an inspector-facing control interface.

Validation of Nondestructive Evaluation Data of RC Bridge Decks

The Validation of Nondestructive Evaluation (NDE) Data of RC Bridge Decks project is a research initiative focused on improving how engineers detect and assess subsurface defects in reinforced concrete bridge decks, without damaging the structures in the process.

The project addresses a critical gap in the field: while AI has the potential to automate defect detection from NDE data, developing reliable AI models requires large, properly annotated datasets. Subsurface defect datasets have historically been scarce, limited to lab specimens and destructive testing methods that don’t reflect real-world conditions. To solve this, the research team built a validated, annotated dataset for three established NDE techniques: Infrared Thermography (IRT), Ground Penetrating Radar (GPR) and Impact Echo (IE). Data was collected from five actual in-service bridges in Grand Forks, N.D. that were already scheduled for repair and maintenance.

Key highlights include:

• Targets a major barrier to AI adoption in bridge inspection: the lack of real-world, labeled subsurface defect data.
• Data collected from five live Grand Forks bridges, making the dataset grounded in real conditions rather than lab simulations.
• Covers three complementary NDE methods for a more complete picture of bridge deck health.
• Lays the groundwork for training AI models that can eventually automate defect detection at scale.

Use of Hyperspectral Imaging for 3D Printing Quality Control

The Hyperspectral Imaging (HSI) for Civil Infrastructure Assessment project is an ongoing research initiative exploring HSI as a unified, noncontact sensing platform capable of detecting subtle surface-level changes in civil infrastructure materials, before those changes become visible or structurally significant. By capturing reflectance data across hundreds of wavelengths simultaneously, HSI reveals chemical and physical surface conditions that traditional nondestructive evaluation methods cannot reliably detect.

The project addresses a recurring challenge across infrastructure inspection: conventional imaging and standard NDE techniques lack the spectral resolution to identify early-stage material changes such as corrosion onset, moisture phase transitions, fouling accumulation and hydration dynamics. To address this, the research team is applying HSI across four distinct infrastructure applications, using dimensionality reduction and change-point detection to isolate the wavelengths most sensitive to each condition of interest.

Key highlights include:

• Early-stage corrosion on structural steel detected before visual onset, with VNIR wavelengths (540–785 nm) producing the strongest spectral response and approximately 90% reflectance change in early corrosion.
• Spectral signatures used to detect and quantify railroad ballast fouling levels, offering a non-contact alternative to labor-intensive physical sampling.
• OH band absorption tracked during early concrete hydration to monitor moisture-driven surface changes in both conventional and 3D-printed concrete (3DCP).
• HSI evaluated for its ability to consistently distinguish wet, slush and ice conditions on pavement surfaces in support of winter road safety monitoring.