Why Self-Driving Cars Are Still a Long Way Down the Road 352
moon_unit2 writes "Technology Review has a piece on the reality behind all the hype surrounding self-driving, or driverless, cars. From the article: 'Vehicle automation is being developed at a blistering pace, and it should make driving safer, more fuel-efficient, and less tiring. But despite such progress and the attention surrounding Google's "self-driving" cars, full autonomy remains a distant destination. A truly autonomous car, one capable of dealing with any real-world situation, would require much smarter artificial intelligence than Google or anyone else has developed. The problem is that until the moment our cars can completely take over, we will need automotive technologies to strike a tricky balance: they will have to extend our abilities without doing too much for the driver.'"
Not about fully automated cars (Score:4, Informative)
The article is about semi-automated cars, not fully automated ones. Semi-automated cars are iffy. We already have this problem with aircraft, where the control systems and the pilot share control. Problems with being in the wrong mode, or incorrectly dealing with some error, come up regularly in accident reports. Pilots of larger aircraft go through elaborate training to instill the correct procedures for such situations. Drivers don't.
A big problem is that the combination of automatic cruise control (the type that can measure the distance to the car ahead) plus lane departure control is enough to create the illusion of automatic driving. Most of the time, that's good enough. But not all the time. Drivers will tend to let the automation drive, even though it's not really enough to handle emergency situations. This will lead to trouble.
On the other hand, the semi-auto systems handle some common accident situations better than humans. In particular, sudden braking of the car ahead is reacted to faster than humans can do it.
The fully automatic driving systems have real situational awareness - they're constantly watching in all directions with lidar, radar, and cameras. The semi-auto systems don't have that much information coming in. The Velodyne rotating multibeam LIDAR still costs far too much for commercial deployment. (That's the wrong approach anyway. The Advanced Scientific Concepts flash lidar is the way to go for production. It's way too expensive because it's hand-built and requires custom sensor ICs. Those problems can be fixed.)
Re:Taxis first (Score:5, Informative)
I think we'll probably see self-driving cars in congested, relatively low-speed environments like inner cities before they're screaming down the highway at 75mph.
On the contrary, "screaming down the highway at 75mph" (never been on the Autobahn, have you?) is a lot easier to automate than driving around a city block. Similarly the easier part of a plane's autopilot is the part that handles cruising at 500mph at 30,000 feet. The numbers are impressive, but the control is comparatively easy.
On a highway there are no traffic lights or stop signs, and there are nicely marked lanes and shoulders. Just stay between the lines at a constant speed and hit the brakes if something appears in front. Compare that to trying to figure out if some guy who's not watching is going to step off the curb and into your way, or if the car pulling out of a parking spot is going to wait for you to pass.
Re:Don't have to be perfect, just better (Score:5, Informative)
http://www.census.gov/compendia/statab/cats/transportation/motor_vehicle_accidents_and_fatalities.html [census.gov]
The catch is, nearly all traffic accidents are preventable by one of the parties involved. Most are at low speeds and most are due to the driver not paying attention to the situtation around them. Next time you are at a busy traffic light, count the cars around you. Chances are one of them will be in an accident that year. Now do that every time you stop at a traffic light....