GPU-Supported Real-Time Sensor Modeling for ADAS Development in Simulink

Electronic control units (ECUs) for advanced driver assistance systems strongly rely on sensor data to interpret a complex environment and traffic situations. Today development and testing of such ECUs is mostly supported by simulations applying a virtual vehicle approach in several steps from MIL via SIL to HIL. Since there are countless different traffic situations, simulation of such scenarios can save time and effort over real-world testing while providing better test coverage. During the development and testing of the corresponding algorithms it is important to provide realistic sensor signals for the simulation in real time. In this session, we present an approach for parallelization of the computing processes that enables model equations to be solved more quickly. The overall vehicle model and the ECU algorithm are simulated in Simulink^®. For the physical sensor simulation, the native GPU of a PC is used. We also present benchmark results from development of a parking assist system that uses ultrasonic sensors in a parallel parking scenario. The simulation is executed on a Simulink Real-Time™ and Speedgoat HIL system.