Safety validation of automated driving functions is a major challenge that is partly tackled by means of simulation-based testing. The virtual validation approach always entails the modeling of automotive perception sensors and their environment. In the real world, these sensors are exposed to adverse influences by environmental conditions such as rain, fog, snow, etc. Therefore, such influences need to be reflected in the simulation models. In this publication, a novel data set is introduced and analyzed. This data set contains lidar data with synchronized reference measurements of weather conditions from a stationary long-term experiment. Recorded weather conditions comprise fog, rain, snow, and direct sunlight. The data are analyzed by pairing lidar values, such as the number of detections in the atmosphere, with weather parameters such as rain rate in mm/h. This results in expectation values, which can directly be utilized for stochastic modeling or model calibration and validation. The results show vast differences in the number of atmospheric detections, range distribution, and attenuation between the different sensors of the data set.

Validating safety is an unsolved challenge before autonomous driving on public roads is possible. Since only the use of simulation-based test procedures can lead to an economically viable solution for safety validation, computationally efficient simulation models with validated fidelity are demanded. A central part of the overall simulation tool chain is the simulation of the perception components. In this work, a sequential modular approach for simulation of active perception sensor systems is presented on the example of lidar. It enables the required level of fidelity of synthetic object list data for safety validation using beforehand simulated point clouds. The elaborated framework around the sequential modules provides standardized interfaces packaging for co-simulation such as Open Simulation Interface (OSI) and Functional Mockup Interface (FMI), while providing a new level of modularity, testability, interchangeability, and distributability. The fidelity of the sequential approach is demonstrated on an everyday scenario at an intersection that is performed in reality at first and reproduced in simulation afterwards. The synthetic point cloud is generated by a sensor model with high fidelity and processed by a tracking model afterwards, which, therefore, outputs bounding boxes and trajectories that are close to reality.

Perception sensor modeling is essential for the safety validation of automated driving systems in virtual environments. Nevertheless, the community lacks a methodical approach to derive requirements for such sensor models that enables a serious application for safety validation in the first place. This article provides a method to specify sensor models for the environmental perception of automated driving systems. The key of the approach is a collaborative collection of cause-effect chains as the basis for specification. With this collection at hand, a tabular form is introduced to extract the relevance of the effect chains to be modeled. Combined profound expert assessments in the table enable the test engineer to specify sensor models within a traceable decision-making process.

Sensor and perception simulation is key for simulation based testing of automated driving functions. Depending on the testing use-case, different cause-effect chains for the specific sensor technology become relevant and the demanded computation times differ. In this work, a novel approach for object-based lidar simulation is introduced, identifying and modeling major sensor effects while balancing effect fidelity and computation time. With an explicitly designed static experiment, simple object based models are falsified, showing the need of a novel approach. Therefore, refined bounding boxes are designed and integrated into occlusion calculation. This new approach is compared to an advanced ray tracing simulation on an object output level. The comparison is conducted on the cause-effect chain of partial object occlusion, which has been identified as highly relevant. The modeling approach challenges ray tracing as the magic bullet for high-fidelity lidar front-end simulation. In direct comparison with the ray tracing approach, our novel approach stands out due to its significantly lower computation time. The newly developed object-based model is open source and publicly available at https://gitlab.com/tuda-fzd/perception-sensor-modeling/object-based-generic-perception-object-model .