The discussion around autonomous vehicles (AVs) is chronically fixated on two areas of technology: Sensors and AI algorithms
Even when simulator technology enters the conversation, we tend to focus on advances in using simulation to train AIs. Tesla claims that their edge in developing autonomous vehicle AIs—and rolling out “Automatic driving on city streets” within the year—is that they’ve eliminated the need for simulation technology to train their AIs. (As reported by the Washington Post in July 2019: “[Tesla’s] artificial-intelligence program is being trained in real-time by data collected from every Tesla already on the road. Every touch of the steering wheel helps inform the company’s software of how to react to various scenarios.”)
What’s left out of this conversation is the fact that the real challenge facing ubiquitous AVs isn’t the technology; it’s the people.
We already know that people develop extremely bad habits with autonomous vehicles extremely quickly. In one study nearly one-third of the participants got into crashes owing to “overtrusting” the AI system—failing to take over control, even when the vehicle clearly alerted them to do so. In another study, several participants closed their eyes or fell asleep, despite knowing that they may need to take over driving at any time.
Using Simulator Technology to Understand the Human-Machine Interface
In almost every technological advancement, the human side of the equation is the most challenging to solve. That’s why Realtime Technologies has invested so heavily in developing SimDriver, a dedicated autonomous vehicle simulation software module for its flagship SimCreator DX research-grade simulation development platform.
As Heather Stoner, General Manager for Realtime Technologies has often noted, “If you want to really see what people are going to do once they’re in an autonomous vehicle, they need to feel like they’re actually in a vehicle.”
This means furnishing simulator technology with high-fidelity vehicle dynamics, rich scenario-building capabilities, and the most realistic possible hardware: full cab simulators, big projection screens, force-feedback steering wheels, and so on.
But it also means giving researchers an enormous amount of fine-grained control of what happens in the simulation. With SimDriver, researchers can select a default profile for the entire scenario, or create their own, individually dialing in each element of autonomous vehicle behavior. These profiles can be applied to the ownship, a single vehicle in the simulation, all ambient traffic, or distribution of profiles across all vehicles present in the simulation. They can also specify pedestrian behavior, weather conditions, lighting conditions, even vehicle malfunctions—anything that might throw the occupant of an AV a potentially deadly curveball.
Most importantly, SimDriver is fully integrated into the SimCreator DX suite—where complete, complex, driving scenarios can be completely developed in short order, without writing a line of code. Curious? Check out our quick-start guide to learn more about rapid prototyping in SimDriver DX.