Self-driving cars, we’re hearing about them every single day. It seems the technology continues to develop, and each day we’re a stop closer to having autonomous vehicles on the road. This has sparked many a debate and discussion about safety and security. And now, we’ve got another piece of the puzzle to add to your debate.
These self-driving vehicles are going to learn how to drive not from specific coding or robotics, but from a video game. Grand Theft Auto, at that.
Grand Theft Auto V is a simulation platform that engineers and researchers are using to train and test the machines that will one day take control of your mini-van. Sound frightening? But experts say you shouldn’t be worried.
Davide Bacchet leads the simulation efforts for Nio, a startup that hopes to get an autonomous electric car in 2020. “Just relying on data from the roads is not practical,” he says. “With simulation, you can run the same scenario over and over again for infinite times, then test it again.”
Larger companies like Ford and Waymo may have similar goals to get cars to market in 3 years, but there’s so much to learn about how to respond when unknown situations arise. As crazy as it sounds, these hyper-real video games are able to generate data that’s closer to real life situations, and give the machines the chance for trial and error.
Scientists from Darmstadt University of Technology in Germany and Intel Labs created a method to grab visual info from Grand Theft Auto V last year. Now, that same kind of data is being tweaked for use in the self-driving sector.
Alain Kornhauser, a Princeton University professor who advises the Princeton Autonomous Vehicle Engineering team, said that while the ideas of highways and side streets in a fictional city isn’t a replacement for asphalt, the game “is the richest virtual environment that we could extract data from.”
Waymo is using simulation to create complicated motoring situations for any possible scenario; for example, having three cars changing lanes at the same time at various speeds and directions. What they learn in the virtual realm is applied in the physical realm, and any problems they encounter are evaluated and fixed. “We’re able to play back the exact situation and predict via simulation what could have happened if the car had been left to drive itself,” Waymo said in a self-driving project report. “If the simulator shows better driving is called for, our engineers can make refinements to the software, and run those changes in simulation in order to test the fixes.”
The question yet to be answered is what requirements state and federal safety regulators will need as proof that self-driving cars are OK to have on the streets. Rules have yet to be written, but already companies are working their hardest to be prepared.
So while self-driving cars still may be a few years out, the next time you’re playing Grand Theft Auto, you may be encountering the same kind of problem the latest self-driving car is working on. It’s pretty strange, but pretty awesome that technology is allowing work to be accomplished at this pace, and with such ease.