The development of automated driving systems is increasingly virtualised with the help of simulated vehicle environments. However, perception sensors, as the interface between the vehicle and the open world, pose one of the greatest challenges. Perception sensor simulation software saves our customers' time, money, and test engineers. It transfers most of the testing and safety validation effort from the real world to simulation. Persival supports the use of perception sensor models for each individual use case, from defining requirements, to building custom sensor models, to validating with real-world measurements. Our simulation and validation services are not limited to the automotive industry, as we also provide virtual camera, lidar, radar and ultrasonic sensors for any wheeled and non-wheeled robotic and infrastructure applications.
As a spin-off from the Institute of Automotive Engineering (FZD) at Technical University of Darmstadt, the company's know-how is based on several years of research under the supervision of Prof. Dr. rer. nat. Hermann Winner, who was the supervisor for both founders' PhD theses. The business is a direct result of the research projects SET Level, and VVM on simulation based development and validation of automated driving functions. Together with all major OEMs, TIER1s, and research institutes of the German automotive industry, our two founders have established methods for the definition of sensor model requirements, model development, and model validation based on the earlier projects ENABLE-S3 and PEGASUS. Additionally, we are also a long-time member of the Change Control Board of the ASAM e.V. standard for simulation model interfaces called OpenSimulationInterface (OSI). To advance perception sensor simulation, we strongly support open source projects, e.g. as Sub-Library Maintainer (SLM) of the asc(s e.V. - ENVITED Open Source Model & Simulation Library. Here are some of the key scientific contributions that form the basis of our expertise:
Knowing what you need for your specific use-case is the first major step in building a simulation pipeline.
Next to interfaces, hardware and parameters, selecting sensor effects that have to be modeled is most challenging.
Our founders established the PerCollECT initiative to collect and provide perception sensor cause effect chains in a tree-shaped ontology. Utilizing our Cause, Effect, and Phenomenon Relevance Analysis (CEPRA) we specify sensor models tailored to your use-case with a traceable decision making process.
Perception sensors are exposed to all kinds of adverse environmental conditions in the open world. These conditions range from rain over snow and fog to road spray and exhaust fumes. Especially for optical sensors, e.g. lidar, the influences on the signal propagation lead to range degradation and false positive detections. We model these influences on lidar sensors using stochastics from real world measurements. More details can be found in Clemens' PhD thesis.
We use sample validation methods to compare simulation results with real world measurements. Interpolation and extrapolation of metric results on these samples provides an uncertainty quantification of the model for every operation point of the model. So at every point in your simulation, you know exactly how uncertain the model is. This methodology can also be applied to model calibration, as it allows a multi-dimensional comparison of the simulation model with the real system. More details can be found in Philipp's PhD thesis.
Persival GmbH was founded in 2021 as a spin-off from the Technical University of Darmstadt with the aim of changing the way autonomous systems are developed today. Our sensor simulation products, services and research help our customers to develop safe autonomous systems while saving time and money. Persival GmbH is fully owned by the two founders. This makes us completely independent of any sensor manufacturer, OEM or tool vendor in the market.