On the streets of Mountain View, California drive the oddest cars. Riddled with sensors, they are idolized by young people with smartphones who snap photos to put on their social media profiles. On the bumper of each one is printed, “self-driving car.” They are Google cars, called autonomous cars because they don’t need drivers.
People have envisioned autonomous cars for decades in science fiction novels, as well as having successfully implemented auto-pilot in countless other vehicles. “It’s no accident that autonomy came to other vehicles first,” says the Computer History Museum online. “However distant or exotic, the sea, the air, and even the surface of Mars are relatively forgiving environments for self-guiding vehicles. There are no children to dart out in their path; no traffic lights, or distracting billboards.”
A lot of early planners for autonomous cars decided that highways would need to be the containers for any car that would drive itself. GM and RCA in 1953 created a prototype of a highway where steel cabling underlaid the road. The cars were propelled by magnets. This design for autonomous vehicles was obviously never widely adopted due to infrastructure requirements to put steel cabling beneath every public road.
Fast-forward to 2004, when autonomous vehicle research really got interesting. U.S. Defense Advanced Research Projects Administration (DARPA) promised one million dollars to the team that could drive their car across a route in the Mojave Desert. Fifteen teams qualified and raced in the first DARPA Grand Challenge.
“The Grand Challenge proved to be one of the more humbling events in automotive history,” wrote Burkhard Bilger for The New Yorker. “Its sole consolation lay in shared misery.”
That year, none of the teams finished the course. In fact, none of them even made it eight miles. The next year was less of a disaster. Five vehicles successfully completed the course.
“Prime goals remain safety, speed, access, more cars sharing the road, intelligent intersections and reducing congestion,” says the Computer History Museum. All of those things come together for a hefty bill. The reason no vehicles completed the DARPA 2004 course is that self-driving cars are remarkably difficult to get right.
“Ninety-eight percent of driving is just following the dotted line. It’s the other two percent that matters,” wrote Bilger.
In an episode of the show Silicon Valley, Jared is picked up by a self-driving car. He happily gets in and tells the car his desired destination. Halfway through the drive, the car receives instructions that it is supposed to be shipped to the owner’s private island, so it turns around and drives to a loading dock and into a shipping crate, where Jared gets locked inside for the duration of the 103-hour journey.
How They Work
The autonomous cars being developed today require three main sections: a GPS system, a system to process dynamic road conditions and a system to drive the car.
The GPS system would provide maps and navigation for the car. Most modern vehicles are already equipped with this.
The second component, processing unplanned road conditions, is a lot more complicated. It requires radar, cameras and lasers that provide information not contained in the GPS maps. This might include pedestrians, objects in the road, traffic lights and road signs.
“If you think of the map as having a static view of the world, the sensor system is providing a dynamic fill-in to that map,” says Sridhar Lakshmanan, an engineering professor at the University of Michigan-Dearborn. “These two, together, provide what is called a ‘world model’ for that autonomous vehicle.”
The camera lets the car see what is around it. Radar lets the car see up to 100 meters away in the dark or when weather hinders regular vision. Lasers scan the world around in all directions. And advanced algorithms must be employed for all of these things to work together.
The third piece of the self-driving car, the part that does the driving, turning, braking and more, is also available in modern cars. The CAN bus can be accessed in the OBD-II port of most vehicles.
The CAN (controller area network) bus is a standard for vehicles that allows messages to be sent throughout the car to control all aspects of it. If you know how to use it, the CAN bus allows you to hit the gas, apply the brakes, control the radio, adjust your seats and much more.
In Google’s promotional video, some of the cars don’t even have a steering wheel. Test vehicles have driven over one million road miles and have carried regular ol’ people. Google advertises on their website (google.com/selfdrivingcar) that their cars are “designed for riding, not for driving,” although the cars do occasionally ask the driver to take the wheel in uncertain conditions.
Google cars have had accidents – twelve total in the six years they’ve worked on the project. According to Google, not one of the accidents has been the fault of the self-driving car. They now publish monthly accident reports on their website.
Imagine commuting to work in your own autonomous car, catching up on emails or reading the paper with a coffee. It sounds too good to be true, but the technology is sound. The future science fiction writers predicted may be closer than we think.