In a presentation for the Institute for Pure & Applied Mathematics, Litton Industries John M. Leonis Distinguished Associate Professor of ECE Jonathan Sprinkle discussed the challenges and advantages of simulating fleets of autonomous vehicles.
"It's not just getting us from point A to point B," Sprinkle said in the presentation. "It's the ability to do things like make traffic better for everyone if only a small number of vehicles in our traffic flow are doing what we want them to do. Even though there's this really great result from doing something with a large-scale physical system, this outcome depended in large part on simulation that we could do in order to be able to know that it's going to be safe."
According to the study abstract, to test multi-vehicle behaviors, simulations must use more than just position estimates and the ability to set instantaneous acceleration or velocity values. High-fidelity simulation environments enable simulation of the vehicle's motion as well as of sensor behaviors: This permits software-in-the-loop evaluation of the behavior of candidate software. However, these simulation environments do not scale to even tens of vehicles, and simulations exhibit divergent trajectories from the same initial conditions, due to the number of components are added to the system.