Books written by Ray Sullivan

Sunday, 17 February 2013

Driving Me Round the Bend

I've discussed self drive cars before - Google have been trialling a self-driving car for a couple of years and have recently applied for a driver's license for it from the State of Nevada.  There's been a number of other attempts, too, and we're all aware of those cars that can parallel park themselves.  Judging by the view from my office window, there aren't too many in circulation right now.

The attraction of designing a viable self drive vehicle is obvious - humans are pretty adaptable with technology and it will take a very advanced computer to improve on the anticipation skills we all see and utilise daily on our busy roads, but when they do then we may see a real decline in the carnage that unfortunately we still see on our roads from time to time.  Because we are imperfect, we do make mistakes driving, we really can't anticipate every manoeuvre on the road and, of course, we get frustrated with delays and diversions.

So a computer that can operate a car as effectively as a human but is also capable of taking in information about the surroundings as well as about the route ahead should make our roads safer, especially if all cars are talking to a master network.

The mechanics of making a car's controls respond to inputs from a computer model seem to be pretty well cracked, as the parallel parking examples demonstrate.  They also demonstrate a certain capability to monitor the external environment.  But the aim for all the research is for the car to carry out complete trips unaided.

A British university, Oxford, has been testing their concept for a self drive car and like Google are hoping to get permission for it to be allowed on British roads.  Their approach is quite different from Google's though.  They are developing a car that learns, so it is useful for repetitive driving tasks such as the school run or the annual trip to the family holiday destination.  When it doesn't know the route, it passes control back to the driver, otherwise it advises when it is content to take over the driving role.

The development team don't use GPS, the technology that is core to most self drive vehicles, but instead use a learning algorithm and external sensors.  While I don't think this is the answer, there is good reason to shun GPS.  Although GPS is very accurate in military hands - they can pinpoint a position to within a metre which is very handy when you are delivering a $1M Hellfire missile at an adversary, but the commercial version is  a lot less accurate - deliberately so and liable to be turned off in the event of a major war.  In the words of a recent BBC article on commercial GPS it is 'capable of accuracy to within 15 metres in a field, much less accurate in built up areas.'  Come on, if you've ended up in a field, it isn't that accurate.

But a European research team has found a way to augment GPS using inertial sensors so that it is close to military accuracy, and it is believed that the next generation of commercial GPS units will incorporate this hybrid approach and make Sat Nav units more accurate.  If the learning skills being developed by Oxford are merged with the hybrid GPS and whatever Google have been building then we may start to see the first really usable self-drive car.

After all, most of us don't want a car that does all the driving.  Just the return trip from the pub.

If you want to read a novel where a Sat Nav goes rogue with devastating results then why not take a look at Digital Life Form


I can be followed on Twitter - @RayASullivan
or on Facebook - use to find me

Why not take a look at my books and read up on my Biog here

Want to see what B L O'Feld is up to?  Take a look at his website here

Worried/Interested in the secretive world of DLFs?  Take a look at this website dedicated to DLFs here, if you dare!

No comments:

Post a Comment