At Google headquarters in Mountain View, California, computer scientist Sebastian Thrun and his team of software engineers are creating a fleet of self-driving cars. His innovative approach to artificial intelligence is what makes these cars such a success. “Science of Innovation” is produced in partnership with the National Science Foundation and the United States Patent and Trademark Office.
Science of Innovation - Self-Driving Cars
ANN CURRY, reporting:
It's one of the most dangerous things we can do each day-getting into a car as either a driver or a passenger. According to the National Highway Traffic Safety Administration, 2.2 million people were injured in car accidents in the U.S. in 2011 alone. It's statistics like these that motivate Sebastian Thrun to build a safer car.
SEBASTIAN THRUN (Google Fellow): It’s just unacceptable to me. I mean, if you want to be innovative, you have to be unhappy, right? So I am really unhappy about the state of transportation today, and I really want to change it.
CURRY: Thrun is a computer scientist at Stanford University who has received research funding from the National Science Foundation. He is also a Google fellow, an honor granted to outstanding engineers. At Google headquarters in Mountain View, California, Thrun and his team of software engineers are working to create a fleet of safer self-driving cars using artificial intelligence.
THRUN: In the self-driving car there’s this vision that we can equally transform society and make cars safer, right. It’s a really, really big vision.
CURRY: Artificial intelligence is about understanding the mechanisms that underlie human thought and behavior and applying these principles to computing devices, such as Deep Blue the chess playing computer or iRobots used by the military for things like disarming bombs. These intelligent machines are programmed to make decisions based on information they gather from the world around them, much as humans do today.
THRUN: So we are able to think, we are able to make decisions. And we want computers to make equally good or even better decisions.
CURRY: Self-driving cars gather information from multiple sources. A spinning laser range finder on top of the car uses beams of light to detect objects in 360 degrees. A radar on the front bumper determines the range, speed, and direction of objects at close range. Two video cameras on the front dash use a narrow view lens to detect things like traffic lights and stop signs, and a wide view lens to record a video of the entire journey. There is also GPS to help the car locate itself and navigate toward its destination. All of these data gathering devices are connected to a central computer that processes the information and controls the car.
THRUN: It’s able to process all these data streams in real time and turn them into relatively simple things, in driving, the use of hands for the steering wheel, the use of feet for breaking and gas. And that’s about the level of complexity of what comes out of the computer.
CURRY: What makes the self-driving car so innovative isn't the way it gathers information, it's the way the car interprets it. The car's computer is programmed to doubt the information it gathers and second guess its decisions.
NATHANIEL FAIRFIELD (Software Engineer, Google): Figuring out that maybe, you know, that the camera is being blinded by the sun at the moment so it should rely on the laser better or something like that.
CURRY: Nathaniel Fairfield is part of Thrun's team of software engineers. His job is to program the car's computer software to make safer decisions.
FAIRFIELD: Sometimes when people are thinking about how the car works inside, they sort of imagine this huge decision tree, they call it, where it’s like if it’s Monday and it’s before nine o’clock and there is a pedestrian right there and I’m going 55 and there is a car up ahead, then I should “ba da da da da da” right? It’s not like that.
CURRY: Instead, the self-driving car learns through practice: first by mapping the surrounding road while a human is driving, and then combining that information with the data it receives when the car is driving itself.
FAIRFIELD: Once it knows sort of what the local situation is, sort of the tactical situation, and it knows where all the moving objects are around it, it can then make decisions about how it wants to actually steer.
CURRY: Some of the robotic and artificial intelligence systems incorporated into the self-driving vehicle have been developed thanks to NSF funding, including those Thrun developed at Stanford. While a trained driver must monitor the car's decisions on the road, in the future, Thrun says that won't be necessary.
THRUN: It turns out the average American worker spends about an hour a day in commuter traffic. What if he could actually reduce that time and give people the ability to do something else like sleep or already start work?
CURRY: In all, Thrun has received 4 patents from the US Patent and Trademark Office for elements of the self-driving car. And none of them have to do with the car itself. Instead, the patents relate to the car's decision making system and the way it communicates with the occupant. Thrun says this patented technology could potentially be added to any car, giving it the ability to drive itself, as well.
THRUN: We use patents as a way to make sure that we go forward and we have the legal rights to build and sell and manufacture what we are inventing.
CURRY: These self-driving cars have already driven hundreds of thousands of miles in California and Nevada without a single at fault accident. While they are still years from mass production, state legislatures have passed laws in Florida, Nevada, and California that will allow them on the road. Thrun believes self-driving cars have the potential to change driving forever.
THRUN: My wife is at the point where she says please let the car drive. The car is a better driver than you, Sebastian. And I try really hard.
CURRY: For Thrun and his team, the self-driving car is an innovation that may soon pave the way to safer roads in the future.
PRO: America’s been ready for self-driving cars since the horse-and-buggy days
Are Americans ready for autonomous cars? They have been ready for more than a century.
The horses that pulled buggies were autonomous, capable of finding their way home with little or no help from their drivers. So, autonomous travel is nothing new. It’s just better.
Science of Innovation, Artificial Intelligence, AI, Cars, Driving, Self-Driving, Driverless, Software, Technology, Engineering, Engineer, Computers, Accidents, Traffic, Human, Behavoir, Laser, Range, Radar, GPS, Global Positioning System, Data, Information, Gathering, Computer, Google, Mapping, Innovation, United States Patent and Trademark Office, USPTO, National Science Foundation, NSF