Here’s how the New York Times described it:
A loose housing around the shaft to the steering wheel in the radio car caused the uncertain course as the procession got underway. As John Alexander of the Houdina Company, riding in the second car, applied the radio waves, the directing apparatus attached to the shaft in the other automobile failed to grasp it properly.
As a result the radio car careened from left to right, down Broadway, around Columbus Circle, and south on Fifth Avenue, almost running down two trucks and a milk wagon, which took to the curbs for safety. At Forty-seventh 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. It was at Forty-third Street that a crash into a fire engine was barely averted. The police advised Houdina to postpone his experiments, but after the car had been driven up Broadway, it was once more operated by radio along Central Park drives.6
It seems that, at a minimum, there were mechanisms for steering, starting the car, actuating the throttle pedal and brake pedal, and perhaps clutch and shifting. It’s possible they just left it in first or maybe second gear, though I think they’d need some degree of clutch actuation.
The thing seemed to work generally well enough for a proof of concept, and in the overall scope of autonomous vehicles the American Wonder proved that motors, servos (automatic devices with some form of error-sensing and correction), and similar mechanisms could be used to actuate conventional car controls in place of actual human limbs and hands. If we replace those radio signals from a human with signals from onboard cameras, sensors, and computers, you’ve effectively got the basics of how modern autonomous vehicles are built.
One fascinating footnote to this has to do with the inventor’s name: Houdina. As you probably already noticed, that name is an awful lot like Houdini, as in Harry Houdini, the famous illusionist and escape artist. Houdini was not the sort of person to take guff of any kind, ever, and he felt that Houdina was deliberately using a name that sounded like Houdini for the name of his company, Houdina Radio Control Co. Houdini didn’t seem to care that the man’s name was, in fact, Houdina, and was convinced it was all just some dirty ploy to capitalize on Houdini’s success and name recognition. Guys who escape from chains underwater don’t usually write tersely worded letters, and Houdini instead opted for the more direct method of going to Houdina’s office and trashing the place.
Houdini wrecked some furniture and an electric chandelier, and pitched what must have been a very exciting fit. Houdini was summoned to court regarding the incident, but no one from Houdina showed up, so Houdini got away scot-free with the perfect crime of chandelier damage.
1933: Mechanical Mike Autopilot
Even though at the moment autonomous control for cars is the hot topic of (admittedly geeky) conversation, it’s worth remembering that airplanes have been flying themselves, more or less, for decades. At first glance this may seem counterintuitive—aren’t aircrafts dramatically more complex than cars? How do they routinely employ self-piloting systems that automobile makers are still struggling with?
The answer is pretty evident when you think about it for even a slight moment, and chances are most of you already realized it while reading that last sentence. It mostly has to do with this one indisputable fact: the sky is really big and really empty. Obstacle avoidance isn’t really that pressing a concern in the air. The chances of a cyclist pulling out unexpectedly in front of you in the sky are exceedingly remote; even if one did, they’d have much bigger things to worry about than getting hit by a passing airplane.
It’s sort of counterintuitive, but the air is a pretty forgiving place in which to develop autonomous piloting systems, even when accounting for the fact that if anything goes wrong the equivalent process of pulling over to the side of the road ends in a fireball on the ground. The sky is vast and empty, and that’s why fairly crude devices like the Mechanical Mike Autopilot, the first really practical aircraft autopilot, were so successful.
Autopilots like Mechanical Mike and most of the ones that followed are self-piloting in that they can maintain a set course and heading (that is, the compass direction an aircraft is pointed) and altitude, but they’re not concerned with any real obstacle avoidance, unless you count the ground, which is, admittedly, a pretty significant obstacle.
The first aircraft autopilot system was developed by Sperry in 1912, and in many ways wasn’t that different from the secret of Whitehead’s torpedo. The original autopilot, known as a “gyroscopic stabilizer apparatus,”7 was composed of a pair of gyroscopes, one connected to the heading indicator and one to the attitude indicator and controlling, via hydraulics, the airplane’s elevators and rudder. This early autopilot allowed a plane to fly level on a particular compass heading, freeing the pilot from many tedious hours of constant attention.
The first really significant use of an autopilot system came in 1931 when Wiley Post set a record for flying his Lockheed Vega airplane around the world in less than eight days. Post’s Vega had a more developed (than that original 1931 system, at least) Sperry Autopilot installed, which Post nicknamed “Mechanical Mike.”
For what it did, Mechanical Mike was surprisingly small, being only about 9 inches by 10 inches by 15 inches.8 The box housed two air-driven, 15,000 rpm (revolutions per minute) gyroscopes, one for azimuth/direction and one for lateral control of the airplane. Air-actuated servo valves connected to the gyroscopes hydraulically controlled the aileron, elevator, and rudder of the plane, giving full three-axis control of the aircraft.
Mechanical Mike required no electrical power, being entirely pneumatic, and only weighed seventy pounds. These were significant advantages over other autopilot systems, and Mike proved to be remarkably reliable. Mechanical Mike and other early gyro-based autopilot systems are significant in the development of autonomous vehicles because they represent the very first time a fully mechanical vehicular control system was trusted enough to transport passengers. The wide use of the Mechanical Mike following Post’s trip marked the first mass deployment of a situationally/environmentally reactive autonomous vehicle system, and, with nearly every major aircraft today employing a much more advanced version, represents the most common autonomous vehicle fleet currently in use on Earth.
World War II: Project Pigeon
The desire to have an autonomous system that could not just react to basic input about the environment—like speed and heading—but could actually sense, track, follow, or aim at a particular target, has been around since long before we had the means to produce machines capable of such tasks. That’s why things like Project Pigeon came into existence; sometimes, we didn’t want to wait around until we developed such machines, and instead borrowed the equipment we needed from nature. Even if that equipment was pigeons.
An autonomous vehicle that can track an object and make turns or otherwise modify its course to track an object is, naturally, highly useful in warfare, where so much is concerned with sending moving things along very fast to chase down and slam into other moving things. To pull this off, you need some sort of visual tracking and processing system, which pigeons already have.
The famous behaviorist B. F. Skinner realized this, and during World War II employed specially trained pigeons to steer a missile at his intended target. Skinner developed a nose cone for a missile that contained three windows, each one with a pigeon looking out of it. These pigeons had been trained via operant conditioning9 (basically, rewarded when a desired behavior was performed