Google's self-driving simulator environment helps its test drivers and engineers to easily create new scenarios and identify areas for improvement, allowing its cars to drive over 3 million simulated miles each day.
In the company's monthly report, Google revealed certain elements that have become a vital part in developing the software. These include testing the cars on public roads, using a virtual environment for simulated driving, creating new scenarios in the simulator and predicting future scenarios through simulation.
"Our simulator can also help our test drivers and engineers quickly identify any areas for improvement," says Google. "Each time a test driver takes over from the self-driving car, 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. 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."
Google also notes that on average, its self-driving cars drive 10,000 to 15,000 autonomous miles every week on public streets. The company added that before its cars hit a new location autonomously, the cars are first driven manually in order for Google to create a map. This way, important details in a street environment such as bridges, curb heights, lane lines and the height of traffic signals are all taken into account.
Out of all the gathered data, the technology focuses more on traffic light details wherein minimal differences such as how the traffic lights in Austin are dimmer than those in Mountain View and how the sensors are calibrated in order to view traffic lights more clearly are all being considered.
According to Google, one of the benefits of teaching a computer to drive is the latter's capacity to have great memory and recall. In particular, the simulator was able to help them recall and rerun the miles they have driven and make the driving experience more comfortable than before by using an updated software.
Google currently has 22 Lexus RX450h SUVs that are self-driving on public streets with 15 of them being deployed in Mountain View and 7 in Austin. Likewise, the company also has 33 prototypes wherein 26 are self-driving in Mountain View and seven are in Austin.
"All of this simulated driving requires huge computing power and luckily we're able to call upon Google's data centers for help!" said Google.
