Self-driving cars may not be proven safe for decades: report

Google’s self-driving cars may have driven themselves 1.5 million miles since 2009, but it could take hundreds of years for robot-car makers to prove safety at the existing testing rate, a new study said.

“Given that current traffic fatalities and injuries are rare events compared with vehicle miles traveled, we show that fully autonomous vehicles would have to be driven hundreds of millions of miles and sometimes hundreds of billions of miles to demonstrate their safety in terms of fatalities and injuries,” said the Driving to Safety report from the Rand Corporation. “Under even aggressive testing assumptions, existing fleets would take tens and sometimes hundreds of years to drive these miles — an impossible proposition if the aim is to demonstrate performance prior to releasing them for consumer use.

“Developers of this technology and third-party testers cannot simply drive their way to safety.”

Even if companies develop “innovative methods of demonstrating safety and reliability,” it still may not be possible to prove autonomous cars are safe, the report, released Tuesday, said.

However, the report noted that self-driving cars have the potential to reduce the nation’s 32,000 traffic fatalities, by taking human error out of the equation.

For the report’s authors, the problem with determining robot-car safety lies in the comparatively low rate of road injuries and deaths versus miles driven. Although the 32,000 annual traffic deaths is a large number, it’s small compared to the three trillion miles Americans drive every year. So to find the rate at which self-driving cars crash and cause injuries or death, which would determine whether the vehicles are safe – and safer than humans – virtually astronomical numbers of testing miles would need to be driven, according to the report.

Accelerated testing, virtual testing and simulations, mathematical modeling, behavior testing, pilot studies and “extensive focused testing of hardware and software systems” are possible supplements to test driving, to put self-driving cars on the road sooner than 100 years from now, the report said.

Google driverless car project chief Chris Urmson wrote in Medium in January that Google puts its cars through three million simulated miles per day.

“Our powerful simulator generates thousands of virtual testing scenarios for us,” Urmson wrote. “It executes dozens of variations on situations we’ve encountered in the real world by adjusting parameters such as the position and speed of our vehicle and of other road users around us. This helps us test how our car would have performed under slightly different circumstances — valuable preparation for a public road environment in which fractions of seconds can be of critical importance.”

Time is of the essence for the mass deployment of self-driving cars, wrote innovation-management consultant Chunka Mui in Forbes last week. “Every day by which we accelerate success can be measured in lives saved (more than 100), injuries prevented (more than 6,000), and dollars better spent ($3.5 billion),” Mui wrote. “And, that is just in the United States. Worldwide, the numbers are even more staggering. In addition, millions could have speedier access to more affordable transportation. Congestion, pollution and dependency on oil could be reduced.”

Photo: U.S. Transportation Secretary Anthony Foxx and  Google executive chairman Eric Schmidt listen to Chris Urmson, director of Google’s self-driving car project, discuss the revolutionary vehicle at the Google campus in Mountain View in February 2015. (Karl Mondon/Bay Area News Group)

 

Tags: , , , , , , , ,

 

Share this Post



 
 
 
  • Richard Birecki

    Nothing in science is ever “proven” lol … SOoo looking forward to the driverless car. Here’s an interesting column on how it will affect real estate prices. http://me.richtrek.com/2015/09/why-buying-real-estate-in-los-angeles.html interesting theory, agree or disagree

    • Marienchambers1

      “my Aunnty Kathy Is getting. pay hourly on the laptop 98$.”….!ac214ctwo days ago black.. MkLaren P3 bought after. earning. 18,512 DoIIars,,,it was my. previous month’s check. just a little over.17k DoIIars Last month.3-5 hours job a day.and get weekly payouts..it’s realy the simplest. job I have ever Do.. I Joined This 7 months. ago. and now making over. hourly 87 DoIIars… Learn. More right Here !ac214c:➽:➽:➽➽➽➽ http://GlobalSuperJobsReportsEmploymentsEliteGetPayHourly98$…. .❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦:❖❖:❦❦::::::!ac214c…….

  • Kirk Hilles

    Yeah, “proof”. Please “prove” to me how young drivers are qualified to drive from our pathetic driving test (compare to others like in German). Please “prove” how people are better drivers while texting. Please “prove” how the idiots that tailgate and go from lane to lane are safer drivers than technology driven? What’s that, they don’t have to “prove” anything, but technology driven does? Uh huh. Yeah, 1 accident (by a bus driver driving with a sandwich in his hand) in 1.4 million miles doesn’t “prove” that it works. Its just a coincidence.

  • Izumi Laryukov

    I believe we are again hitting a “steep shear” in what current science and technology believes is possible when thinking in a linear sense. Our information processing is exploding in an exponential manner, doubling it’s power every two years. The cognitive dissonance is demonstrated when experts try to extrapolate a given ability from today going forward five, three or even ONE year into the future.

    Here is a historical example from the fairly recent past…

    “Some perspective courtesy of Raymond Kurzweil’s “The singularity is near”. Recall the human genome project. It began in 1993. In 1997 roughly 1 percent of the human genome had been sequenced. Scientists said at the rate they were going it would take 700 years to finish the human genome sequencing project. Based on existing technology in 1997 they were correct. Yet the human genome project was fully completed by 2005. What happened was a failure to understand the impact of the exponential improvement of information processing technology.”

    And almost literally from yesterday…

    When the AlphaGo AI won the 4-1 Go match against Lee Sedol, the experts at AI, the scientists and engineers most intimately associated with development of AI and deep learning were the most astonished. They truly believed, based on their understanding and work with practical applications of AI that such a victory could not be achieved for at LEAST 10 years.

    In THEORY it would take a billion years, but in FACT it will be ready in less than five.
    .

  • Izumi Laryukov

    I believe we are again hitting a “steep shear” in what current science and technology believes is possible when thinking in a linear sense. Our information processing is exploding in an exponential manner, doubling it’s power every two years. The cognitive dissonance is demonstrated when experts try to extrapolate a given ability from today going forward five, three or even ONE year into the future.

    Here is a historical example from the fairly recent past…

    “Some perspective courtesy of Raymond Kurzweil’s “The Singularity Is Near”: Recall the human genome project. It began in 1993. In 1997 roughly 1 percent of the human genome had been sequenced. Scientists said at the rate they were going it would take 700 years to finish the human genome sequencing project. Based on existing technology in 1997 they were correct. Yet the human genome project was fully completed by 2005. What happened was a failure to understand the impact of the exponential improvement of information processing technology.”

    And almost literally from yesterday…

    “When the AlphaGo AI won the 4-1 Go match against Lee Sedol, the experts at AI, the scientists and engineers most intimately associated with development of AI and deep learning were the most astonished. They truly believed, based on their understanding and work with practical applications of AI that such a victory could not be achieved for at LEAST 10 years.”

    In THEORY it would take a hundred years of testing, but in FACT it will be ready in less than five.
    .

 
 
css.php