Challenges
Safe perception of the environment and accurate localization of the ego perspective within the environment are two crucial prerequisites for the virtual Validation of automated driving vehicles. The BMWK-funded joint project RepliCar (reference sensor technology for high-precision sensor Validation for automated driving) is recording a highly accurate representation of reality as a comparative value for the Validation.
Solution
RepliCar aims to integrate sensory ground truth data directly into the simulative Validation. This includes data acquisition, data post-processing, feature extraction, and data fusion through to the object list. The project will rely on proven ASAM standards for data artifacts and interfaces and investigate their interoperability within an end-to-end toolchain for Homologation. A GAIA-X-related data space enables partners to trustfully exchange information while keeping their intellectual property.
Key Benefits
Replicar is a research project to improve processes of virtual Validation of vehicles based on real world data. Explicitly, creating a new level of high-resolution ground truth supports investigating how test procedures must be adapted to validate the sensors' reliability. This is a necessity for Homologation and avoidance of perception and display errors in the sensors.
To enable "sensor-in-the-loop" approaches or simulative Validation for the environment detection sensors, it is necessary to capture highly accurate the “ground truth” of the environment to be simulated (other traffic objects, but also the static environment).
To achieve the effective and efficient Validation of sensors, the sensors and data processing chains used must be validated to the highest requirements in precision and fidelity.
To address this challenge the partners of the BMWK-funded joint project RepliCar aim to develop a reference system with integrated high-resolution radar, camera, lidar, GNSS, and inertial sensors. This reference system will be installed into a testing vehicle. By using particularly high-resolution sensors and a particularly powerful sensor data fusion for object recognition, a highly accurate representation of reality is recorded.
The project partner ANavS GmbH will be developing sensor data fusion for environment perception as well as for self-localization of the ego vehicle. The Institute for High-Frequency Technology and Electronics (IHE) at the Karlsruhe Institute of Technology (KIT), Offenburg University of Applied Sciences (HSO), and Freudenberg FST GmbH are developing the high-precision radar system. This essential element in the reference system is validated by the Institute for Regulation and Control Systems (IRS) at KIT. The project partners AKKA Industry Consulting GmbH, the FZI Research Center for Information Technology, IAVF Antriebstechnik GmbH, IPG Automotive GmbH, and GTÜ Gesellschaft für Technische Überwachung mbH will define and implement all steps from simulative Validation to the Certification the of sensors and functions. The driving demonstrator is being realized by Dr. Ing. h.c. F. Porsche AG, Stuttgart. It records initial data and uses it for the release process of an exemplary perception system.
The individual innovations in the Validation and testing process are based on ASAM standards. This includes the integration of artificial intelligence methods for analyzing recorded scenarios, a "sensor in the loop" test bench, integration into existing Simulation tools, and modular Verification and Validation processes. To allow data sharing between partners, HighQSoft GmbH and RA Consulting GmbH will design a GAIA-X-based data space based on ASAM ODS and other descriptive ASAM standards. The targeted end-point-2-edge-2-cloud architecture includes data capturing in the vehicle (endpoint), data consolidation in a telematics server (edge) and the data space (cloud).
Next to the main challenge of designing a system that records ground truth on an unprecedented scale, the project must provide a solution that enables interoperability and seamless exchange of information across toolchains for research and development in automotive electronics engineering. This task requires the definition and implementation of a reference architecture incorporating the newer Simulation standards such as ASAM OpenX with other ASAM standardization domains and applying extensions for interoperability where needed.
One particular focus of the project is on Telematics connectivity, which is one of the essential prerequisites for the efficient and economical performance of driving tests in the Real Driving Validation (RDV) process. The proposed reference architecture (figure 1) shows the standards-based telematics architecture for the communication technology vehicle connection, which combines existing and industry-proven ASAM standards (e.g. ASAM MCD-1 XCP, ASAM XIL, ASAM MDC-2 D, ASAM MDF, ASAM ODS, ASAM OTX Extensions ...) and newer ASAM standards (e.g. ASAM OpenX, ASAM SOVD, ASAM iLinkRT, ...) as well as other established industry standards (e.g. MQTT, ROS), where the project focuses on the industry standards ROS2, which supports Real-Time applications.
Next to using well established ASAM standards like ASAM MDF or ASAM OpenSCENARIO XML, also recently published standards like ASAM OpenSCENARIO DSL were integrated into the project. Moreover, new developments like the extension of ASAM MDF for camera, radar and lidar data are evaluated and an exchange with the associated standardization groups was established.
Since the research project requires project partners to collaborate and share data artifacts, the methodology of data spaces is introduced alongside the standardization efforts. The key challenge is to establish trust within a technical framework that allows participating partners in a data space to keep their intellectual property while sharing specific information based on transparent contracts. For example, one use case is sharing and executing a scenario extraction service of one partner in the edge node of the partner who owns the recorded rosbag files. This use case requires both, certified and standardized processes and interfaces.
This project shows the potential of the ASAM standards to accelerate engineering for Software-Defined Vehicles and Mobility.
Bolin Zhou, CATARC
The project develops a reference system for ground truth data to improve the whole Validation toolchain and the Homologation process. Also, based on a reference architecture, the interoperability of ASAM standards and GAIA-X-based data spaces are investigated to provide further details on a standardized digital twin Validation process that is interoperable with the domain of automotive electronics engineering.
Such an integrated Standard based architecture will
Real Driving Validation (RDV) is a research project funded by the BMWK on the basis of a decision by the German Bundestag.