Design and Development of a Co-Simulation Framework for Testing Longitudinal ADAS on Heavy-Duty Vehicles

Emissions reduction and safety are two of the main objectives to be ensured when traveling on the road; for this reason, the exploitation of new ADAS (“Advanced Driver Assistance Systems”) technologies can have a strong beneficial impact. Nowadays, more and more assistance systems, such as adaptive cruise control and emergency braking, are present on heavy-duty vehicles to support the driver’s tasks. Therefore, before deployment on a fleet, testing operations must be carried out to assess proper functioning and evaluate performance in critical scenarios. Moreover, safety issues and other aspects, such as vehicle and driver preparation costs, play an important role in project evaluation. Indeed, specialized drivers must be employed to perform on-road tests, and vehicles undergo expensive procedures to install and calibrate the necessary sensors. For these reasons, as a first step, tests cannot be conducted in a real environment, making the creation of a virtual framework a fundamental prerequisite. The objective of this work is to set up a Windows-based co-simulation framework for ADAS testing, connecting different software tools to simulate, in a virtual environment, various traffic scenarios. The heavy-duty vehicle model was also created starting from a passenger car integrating all the required elements. Despite this specific goal, the developed architecture can be exploited for broader purposes. The selected software includes CARLA, used for the virtual environment and vehicle representation, MATLAB/Simulink, used to design the vehicle and ADAS algorithms along with the communication protocol, RoadRunner, for scene and traffic scenario creation, and Python, employed to develop a custom module managing the CARLA world and co-simulation execution. The adopted methodology consisted of several steps. First, the CARLA simulator was compiled, fixing encountered bugs and importing a custom map created in RoadRunner. Then, the Python script was developed to open the world, establish communication with the Simulink vehicle model, spawn vehicles, and control their motion. The trajectories of external vehicles are defined by the traffic scenario, while the ADAS-equipped vehicle/s motion is/are computed in Simulink. To validate the architecture, the vehicle model was developed along with three controller logics and a dedicated communication protocol (the latter to exchange data and ensure synchronization and time alignment between Simulink and CARLA). Finally, different traffic scenarios were created to validate both the framework and the control strategies under various conditions. A further step was performed to test the communication stability and synchronization, as well as the potential scalability of the solution in a distributed model. The Python script and CARLA world were deployed on a different PC, leaving the MATLAB and RoadRunner running on the other laptop: both machines were connected to the same Wi-Fi or Ethernet network to allow the communication between them. The co-simulation ran successfully, with vehicles and traffic scenarios executing synchronously within the CARLA virtual environment.