Moving Beyond Static Mapping
Standard mapping services like OpenStreetMap (OSM) provide topological data—where a path exists—but fail to capture the physical reality of the terrain. For wheelchair users, this creates an "information barrier" where a mapped route may be physically impassable due to steep slopes or surface irregularities. OmniPath addresses this by shifting from passive mapping to proactive environmental auditing, treating accessibility as a data-driven engineering problem rather than a static navigation task.
The Technical Framework
OmniPath functions by synthesizing two primary data sources:
- Network Topology: Leveraging OSM to define the path structure.
- High-Density LiDAR: Utilizing USGS 3DEP data to create a high-fidelity 3D model of the environment.
An agentic framework then traverses this 3D model in 0.5-meter increments. At each interval, the system analyzes three critical physical metrics against ADA compliance standards:
- Running Slope: The grade along the direction of travel.
- Cross Slope: The grade perpendicular to the direction of travel.
- Vertical Discontinuities: Sudden changes in elevation (e.g., curbs or cracks).
These metrics are converted into a weighted severity score, categorizing hazards from "Mild" to "Critical." This allows for the identification of "invisible" barriers that are typically omitted from standard navigation datasets.
Validation and Reliability
To ensure real-world utility, the researchers validated OmniPath against 200 physical ground-truth field surveys conducted at the National Mall. Using stratified random sampling, the system demonstrated significant diagnostic reliability for high-severity hazards, achieving F1-scores of 0.60 for "Severe" and 0.58 for "Critical" categories. This performance confirms that the framework can effectively automate micro-scale inspections, providing a scalable way to audit urban environments for accessibility before a user attempts to navigate them.