Businesses could find commercial applications for self-driving vehicles such as taxi services or the delivery of pizza and other goods.
The Google team has about a dozen self-driving cars in operation — all with a human behind the wheel ready to take over at any time. The cars have driven a combined 300,000 miles in varied traffic conditions without any accidents while under computer control.
“It is very much like cruise control,” Levandowski said. “When you want the machine to drive, it will drive, but when you want, you can grab the steering wheel or press the brake, and the command is directly back in your hands.”
The cars operate by using cameras, radars and lasers to digitize the world and create a three-dimensional model inside the vehicle’s computer memory to tell it what is going on around the car.
“You then can create algorithms that dictate how the computer drives the vehicle,” Levandowski said.
Google has been funding the project on its own, thinking that it will eventually be able to license or sell the technology.
“The business model that emerges out of this is yet to be figured out,” Levandowski said.
The approach is not unlike what it did to create its core search engine. Google figured out later it could be a powerful, revenue-generating advertising format, he said.
There are still many kinks to be worked out. The suite of sensors feeding data to the computer is way too bulky and expensive to fit in a mass-produced car. The vehicles still have trouble mastering snow-covered roads, spotting and understanding temporary construction signals and handling other tricky situations.
The challenge is to make sure the sensors figuring out what the world looks like outside the car don’t get confused.
That’s why General Motors chose New York’s Brooklyn Bridge to test the autonomous functions it is building into its new Cadillac XTS and ATS sedans.