AutoSens 2019, the international automotive sensor conference, honors MMI for its exceptional work in the domain of Automotive & Sensor Technologies highlighting the very best technical excellence driving the automotive industry forward [https://auto-sens.com/awards-winners-2019/].
Honored scope of research and recent research results at MMI
The rising use of Artificial Intelligence (AI) for Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles (AV) significantly increases the need of comprehensive and reproducible system tests and validation. Especially, the functional validation of those systems statistically requires more than 6 billion kilometers to be driven [see Pegasus Project]. With respect to time, costs and test coverage, this cannot be achieved with real test drives alone. It is expected, that the introduction of a virtual test and validation stage based on 3D simulation technology can close this gap.
To allow for freely configurable, automated and systematic system tests of ADAS and AV, we propose the use of Interacting Digital Twins and Virtual Testbeds to analyze a large variety of system setups or operation scenarios on system level. Such Virtual Testbeds are based on e.g. coupled sensor and dynamic simulations and integrate all relevant data processing algorithms involved. Thus, Digital Twins of real sensors generate highly realistic sensor data reducing the gap between simulation and reality while massive-parallel simulations reduce testing times.
Recent research results are: Scalable sensor models for efficient generation of sensor data for training and testing AI . Approaches for making AI more explainable [2, 4]. Automated generation of virtual 3D environments for large-scale functional validation .
- : J. Thieling, M. Mathar, and J. Roßmann, “Automated Generation of Virtual Road Scenarios for Efficient Tests of Driver Assistance Systems”, IEEE AUTOTESTCON 2017
- : J. Thieling and J. Roßmann, “Virtual Automotive Testbeds for Automated and Systematic Tests”, AVS 2018
- : J. Thieling and J. Roßmann, “Highly-Scalable and Generalized Sensor Structures for Efficient Physically-Based Simulation of Multi-Modal Sensor Networks“, IEEE ICST 2018
- : J. Thieling, P. Elspas, and J. Roßmann, “Neural Networks for End-to-End Refinement of Simulated Sensor Data for Automotive Applications“, IEEE SYSCON 2019