ADAS Validation: In continuous long-term tests

quote-icon

It’s not easy to always make the right decisions without a driver. My developers feed me huge amounts of data and constantly subject me to tests with many different tasks, to see if I have learned to get things right.

HARRI

HARRI is on the road to level 3 automated driving. This is an enormous step forward. All the driver has to do, when requested by the driving system, is to take control. The camera, LIDAR, ultrasonics and radar supply the necessary data for autonomous movement, thereby replacing the sensory impressions of a driver. They form the basis for the ability to reliably capture the environment and to plan the appropriate driving performance.

 

Machine learning and validation in cloud-based simulation

Initially, HARRI gets to know and differentiate facets of its environment and other road users in virtual driving scenarios—by means of various learning forms and with the aid of artificial intelligence (AI). The goal is to ensure that even in adverse environmental conditions, such as fog, HARRI has the capacity to accurately recognize and assess the situation and the environment, and plan the appropriate driving scenario. This data-driven learning process requires an enormous number of training scenarios and must be constantly tested in a many different ways.

To implement this validation of the autonomous driving feature, our developer team came up with a new methodology for the scenario-based testing based on project PEGASUS*. Originating from a move or maneuver, such as an evasive maneuver, scenarios are derived and synthetic test cases generated. We are already using this test methodology in order to test automated driving features up to the highest level of autonomy, SAE level 5.

Physical validation on the test track

With a view to the Euro NCAP program for new car assessment, the physical validation of vehicles and sensor technology on the test track is also scheduled. In point of fact, our team is already conducting such tests for customers, e.g. for lane departure warning systems (LDW).

Integrated solutions:

  • Virtual test tool chains in accordance with the PEGASUS* methodology
  • Cloud-based testing
  • Methodical test techniques for automated driving features up to SAE level 5
  • Scenario catalogs and criticality assessments for scenario-based testing
  • Tools and testing technology for physical sensor and vehicle validation

 

 

* Project funded by the Federal Ministry for Economic Affairs and Energy for the purpose of establishing generally accepted quality criteria, tools and techniques as well as scenarios and situations for securing highly automated driving features.