However, that data (camera, lidar,…) is passive, driven by human and logged. That means it is not possible to calibrate, test and validate your active algorithms directly, for example, with respect to safety, comfort, and vehicle dynamics/battery performance of automatic braking or lane change maneuvers.
At Siemens, we have developed a testing method trying to incorporate algorithms testing in the loop with real traffic data, called DriveTwin - digital twin of autonomous driving. The concept is somewhat similar to shadow testing, but extends to more focus on vehicle dynamics and performance, where the control algorithm inputs (steering, throttle) are fed to a high fidelity digital twin vehicle dynamics. There you are able to calibrate, test, validate not only perception such as object detection but also safety metrics and others like vehicle performance, comfort, battery… using real sensor data directly.
This is a nice team work from the R&D ADAS team and colleagues at Siemens Digital Industries Software. Hope you enjoy it.