Some baby steps towards better self-driving cars

Imagine you just spent a great night out with some friends.  You have some delicious food, good conversation, and perhaps a few drinks.  You don't worry about driving home, because you know your car will deliver you there safe and sound while you snooze in the back seat.  Such is one of many possible uses for the autonomous car.

I confess that I feel ambivalent about the idea.  Can we ever fully trust a computer to perform the complex cognitive task of driving as well as a human?  Still, it would be awesome to have your Chevy conduct you from errand to errand while you catch up on the news, or to send your Subaru out to pick up your clothes from the cleaners while you're at work.  As a commuter, I'd particularly benefit from the extra time a cybernetic chauffeur would give me each day.

I do think it'll happen eventually, and maybe within a couple of decades.  Google is already test-driving automated cars in California and Nevada.  If the proper sensing and judgment algorithms are perfected, robots may actually outperform human drivers in some respects.   Unfettered by the sluggish speeds of human nerve conduction velocity (mere tens of meters per second), they might have ultrafast reaction times.  If engineered properly, they will have no blind spots, will never misjudge distances, and will be free of distractions.  They will never get drunk.  They might even be able to accurately sense ice or other obstacles on the road and quickly communicate its location to other cars in the area so that they can take appropriate precautions.

University of  Michigan grad student Nick Carlevaris-Bianco has been working on the sensing part, using various lasers and cameras mounted on a Segway robot to help it build an accurate high-resolution map of its surroundings.

This technology must be introduced very carefully and slowly to avoid serious accidents.  The public and insurance companies need time to catch up.  But with some more development in this field, our grandchildren could be driving -- or rather, riding -- on roads that are virtually accident-free.