The Surprisingly Long History of Driverless Cars

As fleets of driverless taxis fan out in cities across the United States and new cars come with an array of autonomous features, it might be surprising to learn that the self-driving car has been in the works for more than a century. Here are some landmark moments in the evolution of driverless cars.

On July 27, 1925, a former U.S. Army engineer named Francis Houdina wowed spectators in New York City with his “American Wonder,” one of the world’s first driverless cars. Every part of the specially equipped Chandler sedan was controlled remotely by radio: throttle, brakes, steering wheel, even the horn. As the “ghost car” made its way down Broadway, Houdina and the car’s operators followed close behind.

Then everything went haywire. The radio controls malfunctioned, and the Chandler started careening wildly down the street. “At 47th Street Houdina lunged for the steering wheel but could not prevent the car from crashing into the fender of an automobile filled with camera men,” The New York Times reported. “It was at 43rd Street that a crash into a fire engine was barely averted.”

Anthony Townsend, a researcher at Cornell Tech’s Urban Tech Hub, can’t help but see similarities between the 1925 failure of the American Wonder and the high-profile malfunctions that still plague driverless cars today. “It’s a pattern that was established a long time ago with these things,” says Townsend, author of Ghost Road: Beyond the Driverless Car. “They’re going to have bugs and breakdowns, just like any other technology.”

At the 1939 World’s Fair in New York, General Motors displayed its captivating “Futurama” exhibition showing self-driving cars zipping along a traffic-free “highway of the future.” In 20 years, GM predicted that cars on the highway would be controlled by electromagnetic guide wires embedded in the pavement. The idea captured America’s imagination, and by the late 1950s, engineers at the Radio Corporation of America (RCA) were close to making it a reality.

“You reach over to your dashboard and push the button marked Electronic Drive,” read an article in Electronic Age, RCA’s quarterly magazine. “Selecting your lane, you settle back to enjoy the ride as your car adjusts itself to the prescribed speed. You may prefer to read or carry on a conversation with your passengers—or even to catch up on your office work.”

On October 10, 1957, the idea was put to the test on a 400-foot strip of highway outside Lincoln, Nebraska. RCA embedded the roadway with electrified guide wires that sent signals to coils on the bumper of a specially wired Chevrolet. The car steered and accelerated without a hitch. The “Nebraska test” was a breakthrough in driverless technology, and U.S. highway officials were intrigued.

There was only one problem—the cost.

“A study was done…looking at the projected cost of the [new] interstate highway system and what it would cost to add this technology [to it], and it would basically add 40 to 50 percent [to] the construction costs, so it was never seriously considered,” Townsend says.

“Shakey the Robot,” created by a team of artificial intelligence pioneers at Stanford Research Institute in the 1960s, has been called the “great-grandfather of self-driving cars” because it was the first robot equipped with computers, cameras and sensors, and programmed with early AI software to move through the world autonomously.

“Shakey was the ‘first autonomous vehicle’ in the sense that [once] it had a destination, [it could] plan its route to that destination and move itself there,” Townsend says.

Armed with a TV camera, an optical range finder and “bump detectors,” Shakey could map its laboratory environment in real time and respond autonomously to different scenarios and obstacles. Today’s self-driving cars are advanced extensions of Shakey’s decades-old sensor and machine learning technology.

The first working prototypes of fully autonomous self-driving cars were built outside of the United States. In 1977, researchers at the Tsukuba Mechanical Engineering Lab in Japan built a car that could read painted white lines on streets and cruise autonomously at 20 miles per hour.

But the real breakthrough happened in Germany in the 1980s, where an aerospace engineer named Ernst Dickmanns took a self-driving Mercedes-Benz van out for a spin on the Autobahn at nearly 55 miles per hour.

Dickmanns and his colleagues at the Universität der Bundeswehr in Munich built an advanced computer guidance system using the latest cameras, sensors and machine learning algorithms. The news of the autonomous Mercedes van caught the attention of Daimler-Benz, which funded Dickmanns’ continued experiments with driverless technology.

In October 1994, two self-driving Mercedes sedans successfully navigated a Parisian highway without a single human intervention. A year later, another car drove autonomously from Germany to Denmark, covering more than 1,050 miles at speeds up to 111 miles per hour. Daimler later lost interest and cut funding, but Dickmanns is widely considered the father of the modern self-driving car.

In the U.S., researchers at Carnegie Mellon University were some of the first to experiment with fully autonomous vehicles. In 1989, the CMU team road-tested the ALVINN—short for Autonomous Land Vehicle in a Neural Network—in a converted Army ambulance. But its most ambitious effort came in the 1990s with a project called “No Hands Across America,” the first attempt to navigate a self-driving car all the way across the United States.

In July 1995, two members of CMU’s Robotics Institute took a modified Pontiac Trans Sport on a 2,850-mile road trip from Pittsburgh to San Diego. More than 98 percent of the record-breaking drive was fully autonomous.

The Defense Advanced Research Projects Agency is the research and development branch of the U.S. military. Created in 1958, DARPA funded some of the earliest research that became the internet, and in the 2000s, it tried to spur the development of autonomous vehicles in America with $1 million in prize money.

“The Pentagon basically said, we want to see who can build a fully self-driving vehicle, and we’re going to set up some tests and give out some prize money,” Townsend explains. “And all the groups that had been working on this in the ’90s—various labs, various companies—all came into that competition.”

In 2004, the first DARPA Grand Challenge attracted 15 teams attempting to race a self-driving vehicle across a demanding 142-mile course through the Mojave Desert. Not a single entrant finished the race—or made it further than 8 miles—so DARPA doubled the prize money in 2005.

The winner of the second DARPA challenge was Stanley, a modified Volkswagen SUV built by a Stanford University research team under the leadership of computer scientist Sebastian Thrun and several industry partners.

No one really knew it at the time, but Google was sniffing around for folks at the DARPA Challenge to run the team that became the Google self-driving car project, Townsend says. That’s where they found Sebastian Thrun.

In 2010, a journalist at The New York Times discovered the existence of Google’s secret project to develop a fleet of self-driving taxis. Over the next few years, Google engineers used modified Priuses and Lexus SUVs to test their cameras and AI software on the busy streets of San Francisco and Los Angeles as well as highways. Then, in 2014, Google unveiled the “Firefly,” a fully autonomous vehicle without a steering wheel or gas and brake pedals—just two passenger seats.

Firefly transported its first passenger on public roads in Austin, Texas, in 2015 before heading to retirement two years later. A year later, Google spun off its self-driving car division as Waymo, a driverless taxi company now operating in multiple U.S. cities.