All Times EDT
Keywords: autonomous vehicles, self-driving cars
Every year, 1.2 million lives are lost to traffic crashes around the world. 94% of these involve human error; in fact, 2 out of 3 people will be involved in a drunk driving crash in their lifetime. We must do better. At Waymo (formerly known as Google's self-driving car project), we are addressing this problem by building the world's most experienced driver. That experience comprises more than 10 million miles of autonomous driving on public roads in over 25 cities and 10 billion miles of simulated driving. Using machine learning and other data-driven techniques, we distill this logged sensor data into the self-driving car's onboard software and replicate it across our fleet. Meanwhile, offboard analysis of this massive dataset allows us to design tests and draw conclusions about the system's safety, using those results to then make safe deployment decisions. This talk will give an overview of our methodology to: (1) train the onboard software to behave intelligently using machine learning; and (2) ensure the safe operation of our system using offboard data science tools. We are currently operating "Waymo One," a commercial ride-hailing service offering self-driving rides to over 1000 public users in the Phoenix Metro area and we are actively expanding to more cities. We are the only company in the world to offer truly self-driving rides (with no one required in the driver's seat).