10 Reproducible Use of Real Sensor Data in the Lab SensFlow leverages real ADAS and AD scenarios for reproducible HIL tests By Henning Kemper, Senior Specialist Editor, IPG Automotive Rare traffic situations, changing software statuses and high sensor data rates make the validation of modern ADAS and AD functions extremely difficult. An edge case recorded once is only valuable when it can be replayed, analyzed and evaluated under reproducible boundary conditions in the lab. This is where SensFlow from IPG Automotive offers support. The modular software suite injects recorded sensor and vehicle bus data deterministically into electronic control units (ECUs) and thus creates a realistic test basis for hardware-in-the-loop environments, without having to perform the respective test case on the road again. Targeted reuse of real-world data Due to the increasing performance of automated driving functions, the effort in their validation rises as well. Sensor setups become more heterogenous, interfaces more complex and the amount of data to process grows. At the same time, safety-relevant situations in real-world traffic occur infrequently or cannot be reproduced as desired. OEMs, suppliers and development teams thus have a clear requirement: They need to be able to feed real measurement data back into the ECU environment in a controlled manner. SensFlow supports this process through a time-consistent reproduction of recorded sensor data and vehicle bus information. In this way, perception, localization and overall functionality can be assessed under controlled laboratory conditions. Deterministic replay with exact timing What matters is not only the replay of data, but also the exact moment at which they arrive at the ECU. SensFlow is designed for the deterministic replay of recorded sensor data streams, including radar, Deterministic replay of recorded sensor and vehicle bus data for reproducible ADAS and AD validation ultrasonic, camera and lidar. In addition, CAN and Automotive Ethernet communication can be reproduced synchronously. Exact time stamp synchronization ensures that all input data are available in the right order and at the right time. SensFlow is therefore ideal for open-loop tests. The function under test receives recorded or specifically provided input data, without needing a closed simulation feedback loop. Perception algorithms, sensor fusion and decision-making logics can hence be evaluated reproducibly. Especially in early development stages, this approach facilitates quick software comparisons, benchmarking and the targeted analysis of individual functions before full system integration becomes necessary. In this context, the synchronized coordination of all data sources is decisive. If raw sensor data, vehicle bus information and image data are reproduced in a time-consistent manner, an exact representation of the recorded driving situation is created at the ECU input. It lays the foundation for reliable lab results and a sound evaluation of software changes. Flexibility for different ECU environments SensFlow’s modular architecture can be adapted to a variety of ECU architectures, sensor layouts and data formats. Therefore, the software supports both individual ADAS functions and complex AD systems with multiple sensor sources and high data rates. Standardized data formats and defined replay processes enable consistent data management. Development teams can Image: © Andrey_Lobachev/Adobe Stock; Editing: IPG Automotive
RkJQdWJsaXNoZXIy MjUzMzQ=