Simulation and its Contribution to evaluate Highly Automated Driving Functions 2019-01-0140
A key criterium for launching autonomous vehicles on real roads is the knowledge of their capability to ensure traffic safety.
In contrast to ADAS, this information is hardly to achieve as the functional scope of an autonomous driving function exceeds by far the ones of ADAS.
As a consequence real world testing alone is not able anymore to cover a sufficiently large test volume.
This assessment problem imposes new requirements on a valid test concept for autonomous driving.
Enabling virtual test domains to contribute reliable test kilometres could represent a possible solution.
In this paper we discuss the feasibility of simulation frameworks to resimulate real world testing in certain scenarios.
We will demonstrate that the contribution to the test volume is incomplete without ground truth information of the vehicle odometry and corresponding environment model.
As a consequence a local variation of the scenario has to consider both, ground truth and odometry with environment model, in order to yield a meaningful degree of confidence for the risk in the resimulated scenario.
Therefore, we first introduce a static representation of traffic scenarios acting as a test case description for an autonomous driving function.
Afterwards the description based on the vehicle odometry and created environment model as well as the description based on the ground truth measured via Differential GPS are resimulated using the same autonomous driving function as applied in the test vehicle.
The generated traces are compared to the corresponding real world data based on a sensitive scenario risk value.
Finally, deviations in the resulting risk contain the validation for the chosen scenario description and applied simulation models as well as the influence of the sensor errors on the resimulation.
Korbinian Groh, Sebastian Wagner, Thomas Kuehbeck, Alois Knoll
BMW Group, Technical University of Munich, BMW Technology Office USA