Transportation

MIT/IEEE-Published Study Falsely Imagines Computing In Robocars Will Emit Lots Of Carbon. Relax, It’s Unlikely.


A study by MIT grad-student Soumya Sudhakar published by IEEE got some press recently for claiming that a large self-driving car fleet in the distant future could cause as much greenhouse-gas emissions as the current global network of computer servers, cited as producing 0.3% of modern emissions.

The study makes poor assumptions, and as such its conclusion is incorrect, but these sorts of studies are often latched onto by the opponents of new technologies due to their confirmation bias, and used as propaganda. Few dig into the details, so I will below. However they turn out not to matter because of the way self-driving EVs will charge. In particular

  1. Self-driving cars can drive themselves to charging stations any time they are idle and power is cheap
  2. Surplus power, mostly from renewables is the cheapest. In fact, sometimes surplus wind power is even “sold” at a negative price — they pay you to take it.

Proposed increased computing power needs

The researchers start by stating that computers in today’s cars draw around 840 watts of power. That’s as much as a typical house, or a few computers in a data center, though it’s a modest fraction an electric sedan driving 60mph, which uses about 15,000 watts while driving. Furthermore, they expect the amount of power needed to grow with time rather than shrink — perhaps to a point that rivals the energy used to move the car. The sedan on city streets is using a fair bit less, so this can start to be an appreciable part of that total power.

Their math suggests that if there are a billion or more of these cars out there, each driving an hour a day, this adds up to a large amount of electricity. If that electricity comes from our modern fossil-fuel powered grid, it could mean a lot of emissions. They do a lot of math on these invalid assumptions to generate a poor result. They attempt to forecast out several decades to produce these numbers, and in one forecast imagine 3 billion cars.

Let’s unpack all the questionable assumptions

  1. It’s true that some prototype cars may draw power like this for their computing. There are also cars that draw far less. They imagine the software and other sensors won’t get more efficient than the prototypes, when in fact they are likely to get vastly more efficient.
  2. The paper even presumes the computational load will grow exponentially and the only salvation for energy use is that AI processors are also getting more efficient exponentially — but not fast enough, and they would need to double in efficiency every year to catch up. The growth path of computation in prototypes is no predictor for future computational needs.
  3. The forecast is based on a 2022 architecture with many cameras constantly processing all images arriving. With other sensors, like LIDARs, which need less computation, you can also avoid a lot of computation from the cameras on regions of the image where one knows nothing is present. That is just one of many tricks, most not yet invented, which will reduce the vision workload — if vision is in fact the approach decades from now.
  4. The imagine all cars going autonomous. This is unlikely to happen for a very long time, though all cars will have some features of this sort, even as driver assist. However, the driver assist functions are already lower power.
  5. They posit a fossil-heavy grid like today, in spite of the massive efforts underway to improve the emissions of that grid. However, as discussed below, these cars will not draw power from the “average” grid and will likely draw almost entirely from surplus solar power, and other forms of surplus power (nuclear, geothermal, some hydro, wind) and rarely from fossil power.
  6. Almost all current AV efforts are electric vehicles, and this is likely to continue. AVs sped up the transition from gasoline powered cars to electric, a big net win for emissions for decades to come. AVs facilitate this switch because the eliminate all existing issues with EVs in the mind of the public. You don’t care about what is under the hood of a taxi, or about its range or recharge time. There is almost never a need to use fossil fuel.

Energy source of self-driving EVs

Today, our grid consists of a mix of fossil fuels, nuclear, solar, wind, hydro and some other sources. Renewables like solar and wind provide power only when the weather dictates. Nuclear plants provide power all the time whether you need it or not. Hydro and fossil power can be changed during the day (sometimes slowly, sometimes quickly) based on expected and real demand. The full grid capacity is used only on hot summer afternoons and evenings to power air conditioning, though that should change in the future through the use of ice-based thermal storage.

Solar power is now the cheapest form of power plant to build in moderately sunny places, however it only provides power in the daylight. It is expected that a ton of solar will be installed to provide that afternoon power. That same solar will generate power from 8am to 3pm as well, though there may not be enough demand to use it. To avoid wasting it, power companies will sell the surplus cheap. And it is electric cars, especially autonomous ones, that will seek to take that bargain — the AVs will even drive themselves to places to plug in during that period. They won’t want to charge during the peak times unless they have to. It is those peak times where fossil power gets most used — we don’t want to use it when there’s enough other power.

Nuclear plants provide power all day, including all night, but people don’t want much power at night. It doesn’t save much to turn down a nuclear plant at night, so that’s another great time for those cars to charge. Cars will do as much of their charging as they can when power is cheap — their owners are not stupid — and that’s when power is surplus, and that’s when power is not fossil based. They’ll do some charging at peak times if they don’t have a choice, but that will be the exception.

Some areas don’t have or are shutting down nuclear. They will need to rely on fossil, wind and hydro at night, plus storage. This does create an emissions risk because night is definitely the most convenient time to charge cars, as most are idle then. Charging can be done more slowly which is cheaper and better on batteries — solar charging from 9:30 to 11:30 and 1:30pm to 3pm, when there is both a solar surplus and more cars are idle, will need to be done faster.

There are also some cars, that due to bad planning or unusual days, will need to charge at 7pm, which is the fossil peak (and price peak) of the day. They will pay a high price for that so it will be their last resort.

I’m not a fan of battery-swap systems for consumer cars, though they can make sense for fleets. With battery swap, you can charge whenever there is surplus power — green and cheap — without worrying about when the cars are on the road. You need to own more battery packs — the ones in the cars and the ones in the charging stations and the ones waiting to swap — but you totally control when you charge.

Another modest negative note is that the most efficient combined-cycle fossil fuel plants are slow to ramp down in response to demand, so the ability of cars to buy surplus power does make that ramp-down problem go away, which might improve the economics of those fuels, but not that much, as the cars again will only buy power if it’s cheap. It also rewards greater efficiency, which is better than punishing that.

Some forecast an all-renewable grid before too long, either with or even without nuclear. Obviously there’s no big emissions problem for electric cars in that case. New efficiencies, storage technologies and demand side management point the way to making that happen.

But even when the rest of the grid still has plenty of fossil on it, the self-driving EVs will be using almost entirely low-emissions power. The author of the paper could have saved herself all the math — it actually barely matters how much energy the computation in the cars take as far as emissions are concerned. It does matter a lot in terms of range, since every MegaJoule used by the computers is one not usable for moving — so they will try to keep that low if they can, but not because they worry about emissions.



READ NEWS SOURCE

This website uses cookies. By continuing to use this site, you accept our use of cookies.