Demo Videos
PolyVerif Framework with Audio
This demo video explains the use model and complete pipeline of the PolyVerif Framework.
There are four AV validation modules to validate AV stack namely Object Detection and perception Validation, AV Control Validation, Localization Validation, and Mission Planning Validation.
The PolyVerif framework initializes the AV stack (Autoware), Simulator (LGSVL Simulator), and the communication channels across all stacks like ROS2 Bridge.
The PolyVerif framework has also integrated the SCENIC Scenario Description language which helps users to generate dynamic scenarios for the test.
The framework starts different data logging modules at various points in the AV stack to collect critical module information during the execution of the test scenarios.
Users can select the test cases from the database which is written in both Python and Scenic (SDL).
The framework then runs the user-selected test cases and collects information at critical points in the AV stack. The PolyVerif framework then calculates valuable metrics of different AV modules and generates detailed reports for each of the validations.
Object Detection Pipeline
This demo video explains about the Object detection validation pipeline of PolyVerif Framework for Autoware AV stack.
The framework components logs all object information like size and position from LG simulator as Ground Truth and object's position from perception node of AV stack's.
The demo starts with explanation of the high level architecture and computed parameters with one scenario. This also has explanation of captured log files, details of reports with every frame comparisons and consolidated report.
Control Validation Pipeline
This demo video explains the AV Control validation pipeline of PolyVerif Framework. The demo helps in validating the impact of AV perception module performance on AV stack’s control mechanism. It explains the complete pipeline architecture and the computed parametric matrices like time-to-collision, control response times to understand the impact of AV perception modules and control responses of AV.
Localization Validation Pipeline
This demo video explains about the Localization Validation pipeline which is capable to test AV stack's localization algorithms in absence of any GPS/IMU sensor information. The demo also explains validation method which are performed for one-to-one mapping from expected location vs. actual location. It also explains the Python API’s role to introduce noises into GPS/IMU sensor.
Mission Planning Validation Pipeline
This demo video explains AV Mission Planning algorithms validation pipeline which involves end to end vehicle trajectory monitoring with collisions and goal position achievement possibilities. With details of a scenario, it explains generated metrics reports and parameters which computes errors and deviations in trajectory following, collisions occurred and deviation from goal position by the AV.