It shows how the number of tests can be reduced to a few relevant scenarios. Algorithms are developed and validated for this purpose. For example, data from public traffic can now be generated from aerial photographs using machine learning and statistical signal processing. The traffic data is checked for similarity and grouped in a specially developed clustering method. From this, relevant scenario types can be derived, which will be specifically examined again at the CARISSMA test site. In the video this is shown by means of an algorithm for prediction and situation interpretation. Research results on efficient parameter estimation for lateral dynamic controllers as well as robust localisation and vehicle condition estimation using inertial sensors are incorporated into the test execution. Both topics are also part of the research activities of this impulse project.
Here you can find the code for data generation by drone as open source framework