The Dawn of Autonomy: From Horseless Carriages to the First Self-Driving Prototypes

On a crisp October afternoon in 1925, a 1926 Chandler touring car rumbled down Broadway in New York City, its cabin empty save for a squat radio receiver bolted to the dashboard and a driver hunched in the rear seat, twisting a remote control that directed the vehicle’s every move. Pedestrians froze mid-step, their wool fedoras tipping back as the car nimbly wove around a rattling delivery wagon, slowed to a crawl for a toddler chasing a ball across the crosswalk, and pulled to the curb with a soft hiss of brakes—all without a single human hand on the wheel. The *New York Times* dubbed it a “phantom motor car,” a marvel of wireless engineering that seemed to yank the future of transportation straight out of the pages of a *Scientific American* serial and onto the cobblestones of Manhattan. For 15 minutes that afternoon, the dream of a driverless car was not a hypothetical debated in engineering labs or science fiction pulp magazines. It was a tangible, rumbling reality, watched by 10,000 gaping onlookers and reported in newspapers across the globe.

The phantom car was the work of Francis P. Houdina, a 32-year-old Dutch immigrant who had spent the previous decade tinkering with radio control systems for military torpedoes. His Chandler was not, by modern definitions, autonomous: it relied on a continuous radio signal from the rear-seat operator, with no on-board perception, decision-making, or feedback loops to navigate obstacles on its own. But Houdina’s stunt was not a parlor trick. It was the latest iteration of a dream that stretched back nearly a century, a dream that had survived countless failed experiments, public ridicule, and the repeated verdict of engineers that self-driving vehicles were technically impossible. To understand the modern autonomous driving boom—the fleets of Waymo robotaxis cruising Phoenix streets, the Tesla Full Self-Driving beta tested by millions of consumers, the long-haul autonomous trucks rolling across Texas highways—we have to start here, in the smoke and clatter of early 20th century experimentation, when the idea of a car that could drive itself was not just a technical challenge, but a cultural obsession.

The roots of autonomous driving lie not in the internal combustion engine, but in the automated carriage experiments of the early 19th century, long before the first gasoline-powered car hit the road. In 1825, a 26-year-old American engineer named John Stevens mounted a small steam engine on a wooden carriage and set it loose on a circular iron track he had built on his family’s estate in Hoboken, New Jersey. The carriage was guided by a raised ridge along the center of the track, which steered the wheels automatically via a simple mechanical linkage. It reached a top speed of 2.4 kilometers per hour, enough to prove that a vehicle could move without a human operator, provided it was confined to a predefined path. Stevens never intended his invention to be a commercial vehicle; he was testing the feasibility of steam-powered rail transport, a precursor to the first commercial steam locomotive that would debut just three years later. But his experiment planted a seed: if a carriage could steer itself on a track, could it be adapted to run on open roads?

That question drove a generation of inventors through the latter half of the 19th century. In 1878, a British engineer named John Isaac Thornycroft built a steam-powered carriage that used a buried cable running along the road to guide its steering, eliminating the need for a raised track. The cable emitted a low-frequency electromagnetic signal that was picked up by a receiver on the carriage, which adjusted the steering angle to stay centered over the cable. Thornycroft demonstrated his carriage on a 3-kilometer stretch of road outside London, where it carried 12 passengers at a steady 9 kilometers per hour. The *Times of London* called it “a glimpse of a future where the horse is obsolete,” but the project never moved beyond the prototype stage: burying cables along every road in Britain would cost an estimated £2 billion (equivalent to £250 billion today), a sum no government or private investor would touch. For the next 20 years, every autonomous vehicle experiment would rely on some form of external guidance: tracks, cables, buried magnets, or later, radio signals from a remote operator. None could navigate open, unmodified roads on their own, because none had the ability to perceive their environment and make decisions based on what they saw.

That changed, in a small way, in 1898, when a 42-year-old Serbian-American inventor named Nikola Tesla took the stage at Madison Square Garden to demonstrate a new invention: a 1.2-meter-long radio-controlled boat he called the “teleautomatom.” The boat was fitted with a small radio receiver that picked up signals from a control box Tesla held on stage; he used the signals to turn the boat’s rudder, start and stop its engine, and even fire a tiny cannon on its deck. The audience of 20,000 gasped as the boat zipped around a shallow pool, dodging obstacles and responding to Tesla’s commands instantly. When a reporter asked him if the technology could be applied to cars, Tesla nodded. “It is practicable,” he told the *New York Journal*. “It is only a matter of time before a car can be guided by a similar mechanism, without a driver at all.” Tesla never built a self-driving car himself—he was too busy fighting Thomas Edison in the War of Currents and developing his ideas for alternating current power transmission—but his demonstration proved that wireless control of moving vehicles was possible, a breakthrough that would inspire a generation of radio engineers to chase the dream of driverless cars.

The first two decades of the 20th century saw a flurry of radio-controlled vehicle experiments, driven by the rapid advancement of vacuum tube technology that made radio signals more reliable and longer-range. In 1903, a Spanish engineer named Leonardo Torres Quevedo built a remote-controlled tricycle that could follow a predefined path using a combination of radio signals and on-board gyroscopes, a design he demonstrated at the Paris Academy of Sciences. In 1921, the U.S. Army tested a remote-controlled supply truck at Aberdeen Proving Ground in Maryland, using it to transport ammunition across rough terrain without exposing soldiers to enemy fire. But these experiments were all limited to controlled environments: test tracks, shallow pools, military proving grounds. None could operate on open public roads, where unpredictable pedestrians, other vehicles, and road debris made remote control impractical, and the idea of a fully autonomous vehicle that could navigate without any human input remained a fringe fantasy.

That changed in 1925, when Houdina’s phantom car rumbled down Broadway. Houdina had spent the previous year modifying the Chandler with a custom radio control system he had designed, using a 100-watt spark-gap transmitter that sent signals to a receiver on the dashboard, which controlled the steering, throttle, and brakes. The rear-seat operator had a periscope to see the road ahead, and a control panel with joysticks to steer and accelerate. The demonstration was a sensation: Houdina repeated the stunt in Boston, Chicago, and Los Angeles over the next year, drawing crowds of thousands each time. He told reporters he planned to build a fleet of autonomous delivery trucks that could operate without drivers, cutting labor costs for shipping companies by 30%. But Houdina’s dream was short-lived. The spark-gap transmitter he used was prone to interference from other radio signals, and the system had no redundancy: if the signal was lost, the car would coast to a stop, or worse, accelerate out of control. In 1927, a demonstration in Detroit went wrong when the signal was jammed by a nearby radio station, and the car plowed into a crowd of spectators, injuring three people. Houdina’s investors pulled out, and he r