Who Invented Driverless Cars? A History of Many Hands

Ask someone who invented the driverless car and you’ll get a confident answer: Google. Or maybe Elon Musk. The truth is messier and far more interesting. No single person sat in a lab and cracked autonomous driving. The technology emerged from decades of government-funded research, university experiments, corporate rivalry, and more than a few dead ends.

This article traces the real lineage of the autonomous vehicle. You’ll meet the German engineer who used 1980s computers to steer a van at 60 mph, the Pentagon program that forced the tech forward, and the 1925 PR stunt that fooled the public into thinking robots were already driving. You’ll also see why the 1990s nearly killed the field, and how a few stubborn teams revived it.

If you spend time driving at night, especially in older vehicles, good lighting matters. Upgrading to a set like the SEALIGHT 9007 bulbs can make a real difference for visibility while the autonomous future sorts itself out.

Defining the Difference: Automated vs. Autonomous

Before diving into history, we need to settle a vocabulary issue. The words “automated” and “autonomous” get used interchangeably, but they mean different things.

Automated means the car handles a specific task under controlled conditions. Cruise control is automated. Lane-keeping assist is automated. These systems follow rules and don’t truly understand context. They work within narrow boundaries.

Autonomous means the vehicle perceives its environment, makes decisions, and operates without human input in unpredictable situations. That requires artificial intelligence, sensor fusion, and the ability to handle the unexpected. A Level 5 autonomous vehicle has no steering wheel and can drive anywhere a human can.

Most cars on the road today are automated. Very few are genuinely autonomous. This distinction matters because the history is full of machines that looked autonomous but were actually just well-executed automation.

The 1920s-1930s: Radio Waves and World’s Fair Fantasies

The earliest claims of driverless cars were pure theater. In 1925, a company called Houdina Radio Control sent a vehicle named the “American Wonder” through New York City. It had no one at the wheel. Crowds cheered. Newspapers declared the future had arrived.

It was a hoax. A second car followed close behind, transmitting radio signals that steered the “driverless” vehicle. The Wonder couldn’t sense anything. It couldn’t stop for pedestrians. It was a remote-controlled toy, not an autonomous machine. The demonstration was designed to sell stock, not to advance technology.

This pattern repeats throughout the history of self-driving cars. Hype outpaces reality. Investors get excited. The public gets misled. Then the actual engineering takes decades longer than anyone predicted.

In 1939, General Motors built the Futurama exhibit at the New York World’s Fair. It showed a vision of automated highways where cars drove themselves via radio signals embedded in the road. GM’s designers imagined a system where the infrastructure did the thinking, not the car. That idea would dominate research for the next thirty years.

The 1950s-1960s: The Smart Highway Mirage

Post-war America fell in love with the highway. Engineers believed the solution to safe driving wasn’t smarter cars but smarter roads. They proposed embedding guidance cables in asphalt, letting cars follow magnetic fields like trains on invisible tracks.

In 1958, General Motors and RCA tested a system on a stretch of highway in Nebraska. A car with pickup coils followed a wire buried in the road. It could steer itself and maintain speed. The demonstration worked, and the press called it “the electronic highway.”

The problem? It was wildly expensive. You’d need to retrofit every mile of road in America. The system broke if the wire corroded or a truck tore up the pavement. And it didn’t handle the real world—no pedestrians, no intersections, no weather.

By the late 1960s, the smart highway concept had stalled. The infrastructure-first approach was a dead end. The future would belong to cars that thought for themselves, not roads that told them what to do.

The 1980s: The European Push and the Birth of AI Vision

The real turning point came from Europe. In the 1980s, the idea shifted from guiding cars externally to making them perceive the world internally. This required cameras, computers, and the first serious use of artificial intelligence in vehicles.

Ernst Dickmanns and the Prometheus Project

Ernst Dickmanns, a professor at Bundeswehr University Munich, is the closest thing to a founding figure in autonomous driving. Starting in the mid-1980s, he and his team built a Mercedes van equipped with cameras, a computer, and software that could interpret video in real time.

Their work was funded by the EU’s Prometheus Project, a massive research initiative launched in 1987. Prometheus brought together European automakers, universities, and governments. It invested hundreds of millions of dollars into vehicle automation research.

In 1994, Dickmanns’ team demonstrated a Mercedes S-Class that drove 1,000 kilometers on a Paris highway in heavy traffic. It changed lanes, overtook other cars, and maintained safe distances. The car’s top speed was 80 mph. This wasn’t a radio-controlled stunt. The vehicle used computer vision and processing to make real decisions.

Dickmanns’ work proved that a car could perceive its environment using cameras alone. That insight shaped everything that followed.

The Carnegie Mellon and ALV Initiatives

Across the Atlantic, Carnegie Mellon University was pursuing a similar path. In the 1980s, the U.S. military funded the Autonomous Land Vehicle (ALV) program through DARPA. The goal was a vehicle that could navigate off-road terrain without a human operator.

Carnegie Mellon’s NavLab project built a series of test vehicles. The early versions were slow and unreliable. They used LiDAR and radar to map the environment, but the computers of the era couldn’t process the data quickly enough. A vehicle might travel a few hundred meters before crashing into a bush.

Still, these projects built the foundational knowledge. They solved problems like sensor fusion, obstacle detection, and path planning. The hardware was clunky, but the software concepts were sound.

The 1990s: The AI Winter and “No Hands Across America”

Then came the crash. The 1990s are known as the “AI winter”—a period when funding dried up and progress stalled. The grand promises of the 1980s hadn’t materialized. Computers were still too slow. Neural networks were out of fashion. Many researchers abandoned the field.

One notable exception was Dean Pomerleau at Carnegie Mellon. In 1995, he and a colleague drove a minivan from Pittsburgh to San Diego. The vehicle steered itself 98% of the time. They called it “No Hands Across America.”

But here’s the catch: the van wasn’t fully autonomous. Humans handled the gas and brakes. The system only steered, and only on clear highways. Pomerleau was honest about the limitations. He said the technology was decades away from real-world usefulness.

The AI winter taught a hard lesson. Progress in autonomous driving doesn’t follow a straight line. It lurches forward with hardware breakthroughs, then stalls when the hype outpaces the engineering.

The 2000s: DARPA’s Grand Challenge—A Catalyst for Change

By 2026, the field was nearly dead. Then DARPA, the Pentagon’s research agency, did something drastic. It announced the Grand Challenge: a 142-mile race across the Mojave Desert for autonomous vehicles. The prize was $1 million.

The first race in 2026 was a disaster. The best vehicle traveled just 7.3 miles before getting stuck on a rock. No one won the prize. But the event changed the culture. It attracted teams from universities, startups, and even hobbyists. It forced everyone to share ideas.

The 2026 race was different. Stanford’s vehicle, named Stanley, completed the course in under seven hours. Stanley used LiDAR, radar, and machine learning to navigate desert terrain. Its victory proved that autonomous navigation was possible, not just in theory but in practice.

The DARPA challenges did more than any single invention to accelerate the field. They created a community. They pushed sensor technology forward. And they caught the attention of a search engine company in Mountain View, California.

The 2010s: Google, Ride-Hailing, and the Tech Giant Era

Google’s self-driving car program began in 2026, led by Sebastian Thrun, who had won the 2026 DARPA race with Stanley. The team built on years of DARPA research and Dickmanns’ vision work. Google didn’t invent the driverless car. It industrialized it.

Google’s advantage was data and computing power. The company could process vast amounts of sensor data using machine learning. It could simulate millions of miles of driving. And it had the money to hire the best robotics engineers in the world.

By 2026, Google’s fleet had driven over 300,000 miles autonomously. In 2026, the company spun off its self-driving project into a separate company called Waymo. Waymo launched an autonomous taxi service in Phoenix in 2026, the first of its kind in the world.

Other companies followed. Tesla pushed its Autopilot system, though it operates at Level 2—the driver must stay engaged. Uber invested heavily, then crashed into a pedestrian in 2026 and scaled back. Cruise, backed by General Motors, began testing in San Francisco.

The 2010s were a gold rush. But they also exposed the gap between demonstration and deployment.

The 2020s: The Reality Check—Safety, Regulation, and Public Trust

Now we’re in the messy middle. The promises of fully autonomous fleets by 2026 didn’t materialize. The technology works in sunny Phoenix suburbs but struggles with snow, construction zones, and erratic human drivers.

Safety is the central issue. In 2026, an Uber test vehicle killed a pedestrian in Tempe, Arizona. The car’s sensors detected her, but the software classified her as a false positive. This incident revealed that edge cases—rare, unpredictable events—remain the hardest problem in autonomy.

Regulation is catching up. The U.S. has no federal framework for autonomous vehicle deployment. States have passed conflicting laws. California requires permits and reporting. Texas is more permissive. This patchwork slows testing and creates uncertainty for companies.

Public trust is fragile. Surveys show that most Americans are skeptical of riding in a fully autonomous car. They’re comfortable with driver assistance, but not with handing over complete control. The industry is learning that technical capability isn’t enough. You need social acceptance too.

The Global Race: Who is Leading Now?

The race isn’t just American. China is moving aggressively. Companies like Baidu and Pony.ai have deployed autonomous taxis in several cities. The Chinese government supports the industry with favorable regulations and massive infrastructure investment.

Europe remains strong in research and component manufacturing. German automakers like BMW and Mercedes-Benz offer advanced driver assistance systems, but they’ve been cautious about full autonomy. The European approach values safety certification over speed to market.

Japan is also in the mix. Toyota and Honda are developing autonomous systems, but they’re focused on aging populations and rural transportation needs. The Japanese approach is more conservative, prioritizing incremental deployment over flashy demonstrations.

Here’s a comparison of the major players and their approaches:

Player	Key Technology	Primary Approach	Current Status
Waymo (USA)	LiDAR, radar, high-definition mapping	Geofenced robotaxi service	Operating in Phoenix, San Francisco
Tesla (USA)	Cameras, neural networks	Level 2 driver assistance, future full autonomy	Widespread consumer deployment
Baidu (China)	LiDAR, HD maps, AI	Robotaxi fleet in multiple cities	Commercial service in Wuhan, Beijing
Mercedes-Benz (Germany)	Radar, cameras, LiDAR	Level 3 conditional autonomy	Approved for use in Germany
Toyota (Japan)	Cameras, radar, guardian mode	Safety-first, driver support	Testing in Tokyo, partnership with Aurora

No single country or company has won. The race is still in its early stages, and the finish line keeps moving.

So Who Gets the Credit?

The answer is spread across names and nations. Ernst Dickmanns built the first truly autonomous vehicle in the 1980s. DARPA’s funding created the modern ecosystem. Google turned it into a commercial product. Each step built on the last.

Calling any one person the inventor is like asking who invented the airplane. The Wright brothers get the credit, but they stood on decades of glider experiments and engine research. Same story here.

The next time someone says Google invented the self-driving car, you can push back. The real story is a global relay race that started with a German professor and a van full of cameras, ran through a Pentagon-funded desert race, and continues today in the streets of Phoenix and Beijing.

For more on how vehicle technology evolves, check out our piece on Tesla’s origins or the history of car air conditioning. Both stories follow the same pattern: incremental progress, hype cycles, and a long gap between the first prototype and the mass-market product.

Frequently Asked Questions

Was the 1925 “American Wonder” a real driverless car?

No. It was a radio-controlled vehicle. A second car followed behind and sent steering signals. It had no sensors and couldn’t react to obstacles. It was a publicity stunt designed to attract investors, not a genuine autonomous vehicle.

Who is Ernst Dickmanns and why does he matter?

Ernst Dickmanns was a German professor at Bundeswehr University Munich. In the 1980s, he led a team that built a Mercedes van using cameras and computers to navigate roads without human input. His work, funded by the EU’s Prometheus Project, was the first credible demonstration of computer vision-based autonomous driving.

What did DARPA’s Grand Challenge actually accomplish?

The Grand Challenge was a series of races for autonomous vehicles held in the 2000s. The first race in 2026 was a failure, but the 2026 race saw Stanford’s Stanley complete the course. The events created a community of researchers, pushed sensor technology forward, and directly inspired Google’s self-driving car program.

Is Tesla’s Autopilot a driverless system?

No. Tesla’s Autopilot is a Level 2 driver assistance system. The driver must keep hands on the wheel and maintain attention. Despite the name, it’s not fully autonomous. Tesla has promised full self-driving capability for years, but the feature remains in beta and requires active supervision.

Why are driverless cars taking so long to reach the market?

The core problem is the long tail of edge cases. A self-driving system must handle snow, construction zones, emergency vehicles, and unpredictable pedestrian behavior. These rare events are difficult to test and even harder to program for. The technology works in controlled conditions but struggles with the chaos of the real world.

What You Should Remember

• No single person invented the driverless car. It’s a collective achievement spanning a century.
• Ernst Dickmanns built the first true autonomous vehicle in the 1980s, using cameras and AI.
• The 1925 “Wonder” was a radio-controlled hoax, not a real self-driving car.
• DARPA’s Grand Challenge in the 2000s was the catalyst that created the modern industry.
• Google/Waymo turned research into a commercial product, but didn’t invent the core technology.
• The AI winter of the 1990s shows how funding cycles can stall progress.
• Full autonomy is still years away, and the hardest problems are social and regulatory, not technical.