C/AV functions (as well as many ADAS functions) are fundamentally different from the traditional control functions in the following way: Traditional control functions are impacted primarily by the behavior of the vehicle where the control is executed (subject vehicle), and the vehicle’s driver (subject driver). CAV functions are impacted by the behavior of the subject vehicle and driver, but also by the behavior of other vehicles, other drivers, and other objects on the road. Correspondingly, the methods used to test CAV functions would also fundamentally be different from the methods used to test traditional control functions: We have to test CAV functions in the context of multiple vehicles, multiple objects, and their environment.
This translates to a need for being able to model a large number of vehicles, along with the environment, at an adequate level of detail and fidelity, to enable Model Based Development of CAVs. Modeling even an individual C/AV poses challenges in the context of modeling the different sensors that might be used on the CAV and executing the entire system in reasonable computing time. The challenges become greatly magnified when the modeling and simulation needs to happen for an entire eco-system of vehicle and other traffic along with the environment. The individual vehicles need to be simulated, but also the information needs to be shared amongst neighboring vehicles. Thus the simulation middleware time and memory increase more than linearly.
Research Objectives, and anticipated Benefits
We intend to explore simulation architectures that would allow for such (co)simulation of a large number of vehicles. Our approach is to combine two separate enabling technologies: ROS 2.0 for data communication using DDS, and the Functional Mockup Interface / Functional Mockup Unit (FMI/FMU) standards for packaging physical system models for co-simulation.
Some of the specific goals of research under this topic would include (but not be limited to):
- To explore the extension of ROS 2.0 (investigated in topic 2 above) as mechanism for information exchange for a Large Scale simulation environment.
- To explore the use of FMI / FMU as mechanism for encapsulating physical system models for use in Large Scale simulation, in conjunction with ROS 2.0
- To develop a Middleware Architecture for Large Scale Simulations of a vehicle eco-system – capable of simulating 100’s of vehicle, that can be fully autonomous, fully manual, or any level of automation in-between on multiple miles of multi-lane traffic. The traffic shall include pedestrians, stationary obstacles, traffic management systems, lane markings on the road, etc.
- To develop needed sensor, communication, traffic management system, and other models as required to simulate the C/AVs.
We expect that this work will be cross-cutting and be aligned with all major TDPs
Research Themes Alignment:
- Dr. Swami Gopalswamy
- Dr. Srikanth Saripalli