Driving forces

Fuel prices continue to climb, steadily increasing the cost of every mile we travel.  Meanwhile, traffic congestion gets worse with each passing year.  And tragically,  car accidents kill around 43,000 people in the US annually. 

What if all of these statistics could be improved with a single technology?

Autonomous driving – driverless cars – may eventually do just that.  Okay, to be fair, the concept involves a family of technologies: video, infra-red, laser and radar sensors; GPS navigation; AI-controlled motion planning; and a variety of mechanical control systems.  But isn’t that what current-day cars are, anyway: an amalgam of systems?

The idea of driverless vehicles has been explored for decades, but it’s only recently that the supporting technologies have reached a sufficient maturity to really be able to capture the attention of the media and the imagination of the public.  I think recent advances are very indicative of how quickly this technology is going to mature during the next few years.

2007 Urban Challenge winner

In 2004, DARPA (Defense Advanced Research Projects Agency) held their first Grand Challenge.  It was over a 150 mile desert course which none of the twenty-one contestants finished.  In fact, the longest distance covered by any of the vehicles was only a little over seven miles.  In the second Challenge held in 2005, five of the vehicles finished a 132 mile off-road course.  All but one of the twenty-three entrants surpassed the prior year’s top distance of 7.36 miles.

Last year, DARPA held their 2007 Urban Challenge.  Of the 11 finalists, six completed the 55 mile urban course, three within the 6 hour time limit.  Rules included obeying all California state driving laws while negotiating with other traffic and obstacles and merging into traffic.  The $2 million prize was won by Tartan Racing, a collaborative effort by Carnegie Mellon University and General Motors Corporation.  Their vehicle, a Chevy Tahoe, covered the course in 4 hours 10 minutes for an average speed of nearly 14 mph.

Last month, GM and Carnegie Mellon announced a new Collaborative Research Lab and a $5 million commitment to work jointly on technologies that will accelerate the emerging field of autonomous driving.  This follows an announcement by GM in January that the company plans to test driverless car technology by 2015 and have cars on the road around 2018.  I wouldn’t be at all surprised if competitive pressures and AI advances moved this forward by a couple of years. 

In the end, regulatory issues and public acceptance of the systems’ safety may delay wide-scale use by several years, but ultimately these vehicles will become the norm.  A properly designed machine can easily react to a detected condition many times faster than a human being.  On-board transmitters and signaling systems could warn of intended actions, giving adjacent vehicles plenty of time to respond.  Combine this with AI swarming algorithms and vehicles will eventually be able to interact with each other with great speed and safety.  (How many collisions have you seen between flocks of birds recently?)

Given the enormous benefits this technology promises (fuel savings, improved utilization of existing roads and lives saved), $5 million dollars seems a trifling sum.  If a more substantial commitment resulted in autonomous vehicles being embraced just one year sooner, how much could we truly save?