As the president of Toyota recently remarked, testing a fully autonomous vehicle is likely to require about 14.2 billion kilometers. As a reference, Google performed a quarter of that distance in six years, using a fleet of 60 vehicles.
If the automotive industry is aiming for fully autonomous vehicles to enter the market in the next few years, simulation will play a crucial role, and certified simulators (that is, that guarantee a proven one-to-one correspondence between the virtual world and the real world) are essential to meet such a target on time.
In the Open Technology Forum at the recent Automotive Testing Expo in Stuttgart, Germany, I delivered a presentation that looked at the validation of vehicle control algorithms through simulation. At TXT Next we advocate the need of novel certified simulation platforms to enable the virtual testing of autonomous vehicles, replacing time-consuming field testing.
The simulator will enable testers to reach targets for test kilometers, which countries are about to set, to enable the road testing of autonomous vehicles, and will provide an acceleration factor enabling hundreds of thousands of test kilometers to be covered a day, with the same level of accuracy of field testing using real vehicles.
A certification procedure needs to be developed, which is cost effective, and recognized and accepted by legislators, enabling performing most of the required road testing in the virtual world, instead of the real world.
Leveraging the experience gained in developing certified simulation for avionic use, we are confident that cost-effective procedures for achieving the certification of existing simulators are possible.