Transportation planning and mobility rely on accurate, detailed data about pedestrian infrastructure, yet traditional collection methods are often labor-intensive and inconsistent. This session explores innovative approaches that combine lidar and AI to modernize how pedestrian pathways and accessibility features are measured, mapped, and analyzed. Topics include using lidar for precise curb ramp assessments to support ADA compliance, advanced AI-driven workflows for large-scale pathway network creation, and deep learning applications that transform high-resolution imagery into actionable feature layers. Attendees will learn how these technologies improve measurement accuracy, reduce human error, enable scalable data collection, and support urban planning and mobility initiatives with rich, connected datasets.
The following presentations will be shared in this session:
Modernizing ADA compliance: Implementing Lidar for Reliable Curb Ramp Assessment
Presented by Michael Olsen, Oregon State University
Ensuring ADA compliance in transportation infrastructure requires accurate, repeatable slope measurements—particularly for curb ramps. However, current field practices relying on digital inclinometers (smart levels) often result in inconsistent and unreliable data. These inaccuracies stem from device-specific limitations, inconsistent measurement practices among inspectors, failure to account for surface flatness, post-construction settlement, and general construction variability. Without standardized measurement tolerances or error models, agencies are forced to reject ramps based on marginal slope deviations, overlooking the inherent imprecision of measurement tools and the natural variability of as-built surfaces. This leads to costly rework and inefficiencies.
Recent research funded by the Oregon Department of Transportation (ODOT) has quantitatively modeled the uncertainty and repeatability of smart level measurements, providing a clearer picture of tool performance under real-world conditions. Additionally, the study evaluated advanced alternatives—including terrestrial, mobile, and handheld lidar systems. These technologies offer key advantages: improved measurement precision, reduced human error, detailed 3D surface characterization, and the creation of as-built digital records that can support broader asset management and compliance workflows. As digital inspection technologies evolve, adopting these data-rich methods represents a critical step toward modern, efficient, and defensible ADA compliance processes.
From Pixels to Planning: Mapping Pedestrian Features with AI
Presented by Rachel Stuckey, Ohio Kentucky Indiana Regional Council of Governments
This project applies a deep learning approach to identify pedestrian infrastructure from high-resolution imagery within the Ohio-Kentucky-Indiana Region. Using ESRI’s pretrained Pedestrian Infrastructure Classification model, the workflow detects features such as crosswalks, sidewalks, and roads. The output raster is transformed into a vector format and further processed with a custom Python script to generate centerlines or representation points of detected crosswalks. This includes the identification of mid-block crossings along functionally classified roads. The final output is a feature layer of point locations representing crosswalks across the region, supporting improved infrastructure planning and spatial analysis.
From Imagery to Graphs with Scalable AI-Driven Pedestrian Pathway and Infrastructure Mapping
Presented by Yuxiang Zhang and Suresh Devalapalli, Gaussian Solutions LLC
Advances in geoscience and AI now enable the automatic creation of large-scale pedestrian pathway datasets, addressing the challenges of traditional methods. Transportation and mobility technologies depend on accurate geospatial data, yet existing datasets are often sparse, inconsistent, and unreliable. This work presents an AI-driven approach to collect and maintain pedestrian pathway data at a statewide scale. By analyzing multimodal inputs that include aerial imagery, streetview imagery and road network data using deep neural network methods, we extract features including sidewalks, footways, crossings, curb ramps, and generate connected pedestrian pathway networks through an end-to-end “human in the loop” AI pipeline. The system employs human validation by surfacing low-confidence predictions and uncertainty zones, enabling efficient review workflows while reducing manual effort. The proposed system significantly outperforms state-of-the-art methods, reducing human intervention and enhancing scalability to large-scale mapping areas. With this solution, we have generated detailed, connected, and routable pathway graphs for the entire Washington state and 4 entire counties in two other states. This work highlights AI’s transformative potential in geospatial sciences, facilitating accurate urban planning and innovative applications in mobility technology.
