What problem does it solve?
Converting recorded driving sensor data (GPS, camera, lidar, actor tracks) into simulation-ready driving scenarios is a complex, error-prone pipeline involving timestamp synchronization, road network acquisition, trajectory smoothing, lane localization, and export to tools like RoadRunner. This Skill guides an AI agent through that entire workflow using Scenario Builder for Automated Driving Toolbox, preventing common mistakes like wrong timestamp scales, missing map origins, and floating actors.
Core Features & Use Cases
- Scenario Generation from Recorded Data: Build ego and actor trajectories from GPS and actor track data, download OpenStreetMap roads, and export to RoadRunner, drivingScenario, OpenSCENARIO, OpenDRIVE, OpenCRG, or Unreal Engine.
- Lane Localization and Validation: Snap ego trajectories to lane centers using camera-based lane detection (RVLD/CLRNet), with mandatory comparison videos and user confirmation gates.
- Specialized Workflows: Extract road surfaces (OpenCRG) from lidar, georeference point clouds, generate 3D assets from images, extract critical events (cut-ins, hard brakes, near-misses), and augment scenes with aerial lidar or traffic signs.
- Use Case: An ADAS engineer has a recorded drive with GPS, camera, and actor tracks and needs a RoadRunner scenario for SOTIF (ISO 21448) coverage. The Skill loads the data, normalizes timestamps, downloads OSM roads, localizes the ego trajectory, exports actors, simulates, and produces a side-by-side validation video.
Quick Start
Ask the agent to generate a RoadRunner scenario from your recorded sensor data file, for example: "Build a driving scenario from the GPS, camera, and actor track data in my sensorData.mat file and export it to RoadRunner."
