• documentation
main content

scenario generation and variation -凯发k8网页登录

generate scenario from recorded real-world sensor data and create scenario variations from seed scenario

in automated driving applications, scenario generation is the process of building virtual scenarios from real-world vehicle data recorded from global positioning system (gps), inertial measurement unit (imu), camera, and lidar sensors. automated driving toolbox™ provides functions and tools to automate scenario generation process. you can preprocess sensor data, extract roads, localize actors and get actor trajectories to create an accurate digital twin of a real-world scenario. simulate the generated scenario and test your automated driving algorithms against real-world data.

to generate scenarios from recorded sensor data, download the scenario builder for automated driving toolbox support package from the add-on explorer. for more information on downloading add-ons, see .

scenario variation is the process of generating multiple variants from a seed scenario that is either manually created or generated from recorded sensor data. you can vary scene parameters, actor parameters, or event parameters of a seed scenario and provide your constraints to generate new scenarios. use these scenario variations for safety assessment of different automated driving applications such as autonomous emergency braking (aeb), lane keep assist (lka), and adaptive cruise control (acc).

to generate scenario variations, download the scenario variant generator for automated driving toolbox support package from the add-on explorer. for more information on downloading add-ons, see .

functions

geographic bounding box coordinates from gps data
extract road properties from road network file or map data
extract properties of roads in path of actor
update lane specifications using sensor detections
generate actor properties from track list
store recorded actor track list data with timestamps
store recorded lane boundary data with timestamps
track lane boundaries
extract properties from input scenario to generate scenario variants
get collision information of colliding actors
store actor and collision property variations
generate variant scenario descriptors from seed scenario descriptor using specified variations
get scenario object from scenario descriptor object

topics

scenario generation


  • learn the basics of generating scenarios from recorded sensor data.

  • format lane detection data to update lane specifications for scenario generation.

  • create jitter-limited ego trajectory by smoothing gps and imu sensor data.

  • localize ego vehicle by fusing gps and imu sensor data to generate virtual driving scenario.

  • perform lane-level localization of ego vehicle using lane detections, hd map data, and gps data.

  • extract lane information from raw camera data to generate asam opendrive® scene or roadrunner scene.

  • extract actor track list from raw camera data for scenario generation.

  • extract actor track list from recorded lidar data using pretrained vehicle detection model and jpda tracker.

  • generate road scene with lanes from labeled camera images and raw lidar data.

  • generate scenario by fusing and smoothing tracked lidar data and camera data.

scenario variation


  • learn the basics of generating scenario variations from a seed scenario.

  • generate scenario variants from seed scenario by modifying actor dimensions.

  • translocate collision from roadrunner seed scenario to target scene.
网站地图