With companies like Google and Volvo working hard to bring self-driving cars to reality, the question of "if" has solidly morphed into one of "when" regarding the technology. Will robotic cars be so prevalent in the next decade or two that our children won't ever need to get driver's licenses?
As part of our regular "One Big Question" series, we put a very similar question to Steven Shladover at the University of California, Berkeley. Shladover is a research engineer who was instrumental in creating California's PATH program (Partners for Advanced Transportation Technologies), whose mission is to "develop solutions that address the challenges of California's surface transportation systems through advanced ideas and technologies and with a focus on greater deployment of those solutions throughout California."
The exact question we put to Shladover and his response follows.
With the advent of driverless cars, will parents today even need to teach their kids how to drive?
Don't expect your children (or even your grandchildren) to grow up without learning how to drive. Although a lot of progress is being made in the development of technologies to automate portions of the driving task under limited conditions, it will be a very, very long time before technology is able to completely replace human drivers under the full range of driving conditions.
Driving is a remarkably complex task that depends on sophisticated perception of the road environment, prediction of the future motions of all the other users sharing that environment (not only other drivers, but also child and adult pedestrians, cyclists and animals) and complex decision making under uncertainty. The complexity of the software that will be needed to implement these functions under the full range of traffic, weather and road conditions far exceeds that of existing software that has been applied in safety-critical conditions. The technology does not exist to design, develop, verify and validate the safety of software of this complexity, and a great deal of fundamental research will be needed to achieve the knowledge required to assure the safety of this software.
None of us would want our children to ride in automated vehicles that are less safe than typical drivers of today, nor would we want to share the road with such vehicles. We need to understand how safe driving is today as a baseline for specifying the safety of automated vehicles.
Based on the U.S. traffic safety statistics, fatal crashes occur an average of once in every 3.3 million hours of driving (representing 375 years of continuous 24/7 driving) and injury crashes occur on average of once in about 65,000 hours of driving (over 7 years of continuous 24/7 driving). Compare those numbers with the average length of time between dropped calls on your mobile phone, or software glitches on your laptop or tablet computer to see how large a gulf remains between today's software and the software that will be needed to safely take over the complete driving task.
We have many opportunities to improve driving comfort, convenience, efficiency and safety through use of automation technology, but those opportunities will for the foreseeable future depend on the automation technology operating in collaboration with a properly skilled human driver for the large majority of applications. There may be some niche applications in which drivers are not needed, but those are likely to be specialized situations such as low-speed urban circulation systems or trucks or buses operating on their own special infrastructure separated from other traffic.
