Power To Run Computers For Autonomous Cars Lowers Range


Here’s a conundrum for you to consider. Auto industry engineers say the computers needed to make autonomous cars possible use so much electrical power, they actually decrease overall efficiency and limit how far an electric car can go on a single battery charge.

Daimler Bosch autonomous cars partnership

We don’t think of the power to run computers all that often. It all happens in the background. Out of sight, out of mind, as the old saying goes. But according to the Lawrence Berkeley National Laboratory, in 2016, 70 billion kilowatt-hours of electricity were needed to run all the servers in the United States.

How much is that? Let’s put it this way. It would take the output from 8 large nuclear power plants or double the amount of energy supplied by all the solar panels installed in the US to provide that much power.

We are enthralled by the prospect of cars that can drive themselves, but there is no free lunch. Engineers at BorgWarner say the amount of computing power required for autonomous cars today is equivalent to having 50 to 100 laptops plugged in and running simultaneously.

“We’ve been battling all the time because the governments are always pushing for a few percent improvement every year,” Scott Gallett, vice president of marketing at BorgWarner, said of fuel-economy standards. “This just amplifies that challenge.”

Chris Thomas, BorgWarner’s chief technology officer estimates a car with Level 4 or Level 5 autonomous driving capability will get 5% to 10% worse fuel economy as a result. That’s a problem in an industry that worries about how many LEDs to use in a brake light because of the impact on fuel economy each one has. “They’re worried about one watt, and now you’re adding a couple thousand,” Thomas said. “It’s not trivial.”

The impact of self driving technology is especially critical at a time when the auto industry and the government are tussling over future fuel economy standards. A fully autonomous Honda Fit will get 54.6 miles to the gallon in 2025 in the best case scenario, which is more than 5 miles per gallon below the U.S. emissions target, according to BorgWarner. A small pickup or SUV would get 45.8 mpg versus a target of 50.

Every one from Elon Musk on down thinks the first use of self driving cars will be for ride hailing and ride sharing fleets. They envision a flotilla of autonomous cars shuttling people from place to place in the world’s overcrowded cities first. Private ownership of Level 5 cars will come later

Mary Gustanski, chief technology officer for Delphi Automotive’s powertrain division, thinks all that computing power on board means manufacturers will need to rely on plug-in hybrids rather than pure battery electric cars, at least until chip manufacturers dramatically reduce the power requirements needed to operate the computers.

Sam Jaffe, founder of Cairn Energy Research Advisors, agrees. “They’re going to favor plug-in hybrid EVs, and they’re going to require that extra gasoline engine, both to extend the range to be able to do a taxi type of duty cycle, but also to help mitigate the proportion of the autonomous systems on the battery pack itself,” said Jaffe, whose research and consulting firm specializes in energy storage.

“If you are trying to maximize your utilization” of an autonomous vehicle, a battery-electric car “is really restrictive for your business,” Jim Farley, Ford’s president of global markets, told investors on Oct. 3. He said Ford believes hybrids are “the right tech to start with.”

Nobody (except Tesla) seems to want to build electric cars, yet every auto manufacturer is pushing aggressively to bring self driving cars to market. Yet those autonomous cars will be a significant drain on overall efficiency and fuel economy, two things that are vital to meeting lower carbon emissions goals. How odd.

It’s a classic case of cognitive dissonance. The technology that will actually reduce emissions is being resisted at every turn. Yet the technology that will make self driving cars possible actually adds a bit to total emission and is being promoted by all parties.

How to resolve the conflict? Easy. Qui bono, as they used to say in ancient Rome, which translates into, “Who benefits?”  That’s another way of saying “Follow the money.” Auto makers foresee enormous profits from operating those fleets of self driving cars but building electric cars as a losing proposition. In the battle between emissions and profits, the outcome is a foregone conclusion.

Source: Bloomberg


About the Author

I have been a car nut since the days when Rob Walker and Henry N. Manney, III graced the pages of Road & Track. Today, I use my trusty Miata for TSD rallies and occasional track days at Lime Rock and Watkins Glen. If it moves on wheels, I'm interested in it. Please follow me on Google + and Twitter.
  • rcbryan

    I hadn’t considered this angle before. Good article.
    I have a few comments to add. All electric cars are developing an ability to self-charge.. ie: wireless or tesla’s “snake”. No such system known for fuel cars. So, autonomous electric vehicles might have to automatically re-charge more often than without autonomous computers. Using Plug Hybrid vehicles as host might preclude autonomous recharge. Also, there may be large enough gains in passenger efficiency and cost to make up for the energy losses [or cost increases] from computers.

  • Ed

    I believe this is at the root of the new Nvidia Drive PX computer, right? It promised huge reductions in power consumption. It is still at 500 watts, I believe, but better than the last generation.

    • Jonny_K

      Holy Moly, 10 trillion ops/second. That’s a lot of computing power. However, when it comes to efficiency, the interesting metric is always an appropriately scaled version of operations/watt-second where operations are some standard operation of interest, not a primitive hardware operation but some relevant computation that is invariant over different hardware and software. I wonder if there is an AI metric as there are for data base operations, web operations etc.

      • Ed

        I don’t know the answer on AI metrics. I was cracking up to realize that my first microprocessor-based production product ran at just over 100kHz! Wow….I wonder how much further this will go while I am still alive?!