With each passing day I grow more pessimistic about the likelihood of ever owning a flying car. The only reasonable people to blame for this dashed dream, of course, are the creators of Back to the Future Part II, who made that phenomenon appear certain—even mundane—in their conception of 2015. (In the flick, regular cars could be converted into hover cars for a measly $39,999.95.) We read stories every week about how society will struggle just to stop using oil by 2015, so I guess fueling our flying DeLoreans with garbage is, for now, out of the question.
To keep up morale in the meantime, an increasing array of goodies has been supplied for drivers who have come to terms with their earth-bound status. DVD players, satellite radio, Bluetooth headsets, GPS and dual climate-control have made car trips, if not enjoyable, bearable. But the same devices that keep our psyches intact in these times of woe can also distract our faculties, turning our attention from the road for a brief but dangerous moment.
For that reason, I ask, when will we have a car that can chauffeur us? When can we sit back in our driver's seat, fall asleep, dream about flying cars and awaken to find ourselves at our destination, snugly parked in a parallel spot?
The idea of autonomous vehicles dates back several thousand years to the ancient Chinese, who are said to have invented the "south-pointing chariot." Thanks to highly advanced gear mechanisms, a figure atop this two-wheeled contraption always pointed south, no matter which direction the chariot traveled. Evidence of these chariots is scarce, but it seems the Chinese did build many of them, and may have used them to raid or flee enemies under cover of darkness. (Today, self-driven vehicles remain of high military interest.)
In the 1970s, the Stanford Research Institute developed "Shakey," considered to be the first artificially intelligent vehicle. Shakey used a television camera to take pictures of several surrounding positions. The cart processed these images for about an hour before deciding where to move, advanced a few feet and repeated the entire process. Not unlike today's Sunday Driver.
But modern cars can also run on auto-pilot, almost as soon as they leave the dealership. Many of these self-operating vehicles faced off recently during the Urban Challenge, a robot race held by the Defense Advanced Research Projects Agency (DARPA). During the 60-mile race, autonomous cars navigated an entire city environment—they accelerated and braked, yielded at intersections and avoided oncoming traffic, all on their own.
Some of the mechanisms aiding this autonomy aren't very advanced. With minor adjustments to the basic chassis, engineers can program a computer inside a vehicle to control its engine, brakes and steering wheel. A series of sensors can read wheel counts and the angle of tires; combined with global-positioning satellites, these devices can estimate where a car is, how fast it's moving and how it can reach a desired location.
Autonomous cars also need external sensors, which include perhaps dozens of lasers, radars and cameras. Some scan upcoming terrain, searching for telephone poles, oncoming cars or octagonal-shaped signs that could impact future actions. Others scan lateral areas, helping robotic vehicles to abide basic rules of four-way stops or even to merge into moving traffic. Some sensors look far ahead, while others focus on nearby obstacles, which can help a car get into, or out of, a tight parking spot.
Add to these functions some accoutrements—automated windshield wipers, cruise control and seat memory, many of which are already available to car consumers—and fully autonomous cars don't seem too far out of reach. To give a sense of the technology's rapid advancement, none of the robotic entrants finished the course when DARPA held its first race in 2004. The following year, four autonomous vehicles completed a course of desert terrain, with few obstacles. This year, half a dozen cars maneuvered through a mock city filled with some 50 human-driven cars, which provided a steady flow of traffic.
But before you spend $39,999.95 to convert your regular car into an auto car, note some drawbacks. Autonomous cars can handle basic traffic, but they can't yet avoid darting objects, such as deer. They don't work well in bad weather. They might navigate Poughkeepsie, New York, just fine, but wouldn't hold up during rush hour in Manhattan.
During the robot race, many vehicles stopped at crosswalks, but often with the help of information embedded in the painted lines. And though they handle four-way stops, artificially intelligent cars can't yet discern colors. In other words, it might detect a traffic light, but it wouldn't know whether to stop, slow down or proceed at pace.
Which raises one final, major question about the future of our automotive industry: If we do one day have flying cars, and with them full-fledged "skyways," where will city, er, aerial planners hang the traffic lights?
The real Wishful Thinker behind this column was Carnegie Mellon robots professor William "Red" Whittaker, whose automated Chevrolet Tahoe, Boss, won the 2007 DARPA Urban Challenge in early November.
Have an idea that should be thought about wishfully? Send it to [email protected]