Test Data Management

Test Data Management

How to Model and Apply Test Data for the Future of Autonomous Driving
The previous chapters looked at the challenges and opportunities of new test strategies in detail, and the role data generation plays in it. By now, the relevance of test data management should be clear. Test data does not appear out of thin air. It must be planned and designed, modeled, and stored, before it can be used. Simply put, the more advanced the test data management strategy, the more efficient the testing of ADAS/AD functions becomes. If defects can be detected and identified early on in the development process they are, naturally, easier to fix. Regarding autonomous driving, high-quality test data management, and hence data exchange and data fusion, is key to facilitating interoperability between systems, manufacturers, technical services, and other stakeholders involved.

Motivation for Investigation of the Test Data Management Aspect
The development of autonomous driving functions can be described as data-driven. In all stages of the validation, huge amounts of already recorded test data, scenarios, and test cases are necessary to investigate the behavior of the function under test (FuT) in different simulations and real-life tests. Therefore, vehicles and test rigs have to collect huge amounts of test data, such as imaging and bus data, while driving on real and simulated roads.

Looking at the importance of having test data and test descriptions available throughout the overall development, simulation, and real testing makes it understandable that key to an efficient development and validation process is an efficient scenario, test case, and test data management.