The virtual testing and development of Advanced Driver Assistance Systems (ADAS) places very similar requirements on our simulation tools to those needed to support Autonomous Vehicle development.
These requirements are:
- The ability to create scenarios with multiple actors (traffic and pedestrians)
- A wide range of sensor models including camera, radar and ultrasound
- Open interfaces to allow the integration of SiL or HiL implementations of the systems under test
- Control of lighting and weather effects
- And the option of having a human driver or automated driver
Using rFpro and the sensor models developed at Claytex we are enabling our customers to make more use of simulation to develop and test these systems.
Let’s look at the detail behind a simulator created for the development and testing of an Automated Emergency Braking (AEB) system. This simulator was built around rFpro and needed to reuse an existing vehicle dynamics model and allow the integration of the AEB controller. The AEB controller was created in Simulink and relies on a forward facing radar sensor. The goal was to provide a simulator that could recreate the NCAP tests and more realistic street scenes.
We configured an rFpro ADAS workstation to run the complete vehicle model (physics and control systems) within the PTWinSim co-simulation environment. This approach allows us to combine multiple apps, compiled from multiple tools, to build up the vehicle model and we can use standard calibration and telemetry tools to interact with the models during the simulation. In this particular case, the control systems were developed using Simulink and the vehicle physics model was built in Dymola.
The diagram below shows the basic architecture of the system, which consists of multiple rFpro instances, performing different tasks and it all runs on a single high performance workstation. The Physics instance is the heart of the simulation and runs the vehicle model and control systems in real-time. The Radar instance hosts the physics-based radar model and the IMU instance hosts the Inertial Measurement Unit (IMU) model. The radar and IMU models output messages directly to the control system using the same message format that they would on the real vehicle. Dedi Server is responsible for synchronising the multiple rFpro instances and controlling the traffic and pedestrians through a traffic plugin.
Our Simulation Manager tool simplifies the task of configuring the multiple rFpro instances needed for this simulation task and allows you to quickly switch between vehicle configurations and test scenarios.
In the test run below we are recreating the Euro NCAP AEB Inter-Urban CCRm test scenario where the vehicle under test (VUT) is travelling at 80 kph with a slower vehicle ahead travelling at 20 kph. In this case a driver model, developed by Claytex, is used to control the VUT so that the test is repeatable. During the simulation we can monitor the response of the control system to make sure that it identifies the slower vehicle and reacts at the appropriate time to avoid a collision.
Using rFpro as our virtual test environment allows us to accurately recreate the test scenario with appropriate sensor models and controller models. The visualisation of the test includes being able to monitor the dashboard warning lights to verify that these activate correctly. The simulation architecture allows us to combine multiple controller models to define the vehicle which allows us to check that the whole system interacts correctly. We can also extend the simulator to add multiple different types of sensors and extend to support HiL testing as required.
Written by: Mike Dempsey – Managing Director