There are a variety of ways to teach driverless cars how to operate on the road, but some scientists are now planning to use one of the most popular gaming franchises, the Grant Theft Auto series, in an effort to teach autonomous technology how to function on streets across the world.

Several research groups are beginning to use the virtual reality world found in Grand Theft Auto to train computer systems that can power driverless cars. Scientists believe that using a technique called machine learning, driverless vehicles can be provided data that is almost as good as real world imagery.

Using a software layer that sits between the game and a computer’s software, labels of cars, pedestrians, and other objects are created and sent to a machine learning algorithm. The system can then begin the process of learning to recognize and identify these different things either on the video game or on a real street. According to the researchers, this way of transmitting information is a lot easier, quicker and cheaper than manually drawing up or creating these same scenarios for learning purposes.

A student who recently coauthored a paper detailing the benefits of using video games for training computer systems discussed the new trend.

“With artificial environments, we can effortlessly gather precisely annotated data at a larger scale with a considerable amount of variation in lighting and climate settings,” said Alireza Shafaei, a Ph.D. student at the University of British Columbia, in an article with MIT Technology Review. “We showed that this synthetic data is almost as good, or sometimes even better, than using real data for training.”

While researches are looking forward to what these games can provide in terms of training computer systems, they do not anticipate that these systems will pick up some of the bad habits, such as speeding and evading police, that are usually associated with the ultra-popular gaming series.

<< Back to the News