Speaker Details

Speaker Company

Martin Herrmann

Martin Herrmann joined IPG Automotive in 2014. In the beginning, he worked on customer-specific simulation and engineering projects with a focus on vehicle dynamics and ADAS. In 2016, he became a business development manager. In this position, he has further expanded his experience and knowledge in the area of ADAS and automated driving, and has been involved in ASAM projects and innovative research projects of the PEGASUS family, for example. Herrmann studied automotive engineering at the Technical University of Ilmenau.


Real-time testing of automated driving functions using parallelization

Autonomous driving functions can no longer be validated solely in the real world, as critical edge cases are rare and dangerous. HIL simulation technologies bridge the gap between simulation and in-vehicle testing. However, the complexity of autonomous vehicles, number of sensors and ever-increasing data rates pose new challenges for HIL simulation. A possible approach is prototypically demonstrated with the implementation of two parallelization concepts based on specific use cases: CPU-based calculations on multiple real-time systems and parallelized sensor simulation on multiple GPUs, allowing arbitrary scalability of the number of sensors as well as the integration of additional models and systems.

The audience will learn:

  • How sensor models with different levels of fidelity contribute to testing automated driving functions in simulation
  • To which degree simulation of raw sensor data can be achieved for HIL testing with the perception layer
  • How the computational load of sensor models can be distributed, observed and optimized on multiple GPUs
  • How the computational load of other simulation modules can be distributed to multiple, time-synchronized real-time systems
  • How the computed raw sensor data (camera, radar, lidar) can be injected into the device under test