Wednesday, March 4, 2015

“What do you mean, I have to learn how not to drive?”

The age of autonomous driving lessons is upon us.

Paul Leroux
What would it be like to ride in an autonomous car? If you were to ask the average Joe, he would likely describe a scenario in which he sips coffee, plays video games, and spends quality time with TSN while the car whisks him to work. The average Jane would, no doubt, provide an equivalent answer. The problem with this scenario is that autonomous doesn’t mean driverless. Until autonomous vehicles become better than humans at handling every potential traffic situation, drivers will have to remain alert much or all of the time, even if their cars do 99.9% of the driving for them.

Otherwise, what happens when a car, faced with a situation it can’t handle, suddenly cedes control to the driver? Or what happens when the car fails to recognize a pedestrian on the road ahead?

Of course, it isn’t easy to maintain a high level of alertness while doing nothing in particular. It takes a certain maturity of mind, or at least a lack of ADD. Which explains why California, a leader in regulations for autonomous vehicles, imposes restrictions on who is allowed to “drive” them. Prerequisites include a near-spotless driving record and more than 10 years without a DUI conviction. Drivers must also complete an autonomous driving program, the length of which depends on the car maker or automotive supplier in question. According to a recent investigation by IEEE Spectrum, Google offers the most comprehensive program — it lasts five weeks and subjects drivers to random checks.

1950s approach to improving driver
alertness. Source:
Modern Mechanix blog

In effect, drivers of autonomous cars have to learn how not to drive. And, as another IEEE article suggests, they may even need a special license.

Ample warnings
Could an autonomous car mitigate the attention issue? Definitely. It could, for example, give the driver ample warning before he or she needs to take over. The forward collision alerts and other informational ADAS functions in the latest QNX technology concept car offer a hint as to how such warnings could operate. For the time being, however, it’s hard to imagine an autonomous car that could always anticipate when it needs to cede control. Until then, informational ADAS will serve as an adjunct, not a replacement, for eyes, ears, and old-fashioned attentiveness.

Nonetheless, research suggests that adaptive cruise control and other technologies that enable autonomous or semi-autonomous driving can, when compared to human drivers, do a better job of avoiding accidents and improving traffic flow. To quote my friend Andy Gryc, autonomous cars would be more “polite” to other vehicles and be better equipped to negotiate inter-vehicle space, enabling more cars to use the same length of road.

Fewer accidents, faster travel times. I could live with that.

2015 approach to improving driver alertness: instrument cluster from the QNX reference vehicle.

No comments:

Post a Comment