Self-driving car computers may be 'as bad' for emissions as datacenters
- Reference: 1673631249
- News link: https://www.theregister.co.uk/2023/01/13/selfdriving_car_computers_could_equal/
- Source link:
This, of course, assumes self-driving cars capable of level 4 or 5 autonomy are [1]actually realized , but the trio of MIT researchers behind the [2]findings say the framework they've built to model carbon emissions from computers inside autonomous vehicles (AV) should call attention to hidden carbon costs and help the auto industry plan for a greener future.
"We are hoping that people will think of emissions and carbon efficiency as important metrics to consider in their designs," [3]said Vivienne Sze, MIT associate professor of electrical engineering and computer science and co-author on the paper.
Our potential AV future is a carbon-heavy one
The MIT researchers say that, if trends in computing power and green energy adoption hold true, and if "projections based on aggressive cost reduction and high customer satisfaction [that] can yield AVs capturing up to 95 percent of the market share by 2050" are correct, an assumed global fleet of one billion AVs driving an average of one hour per day, using a computer that consumes 840w, would consume enough power to equal global datacenter emissions from 2018.
Per the paper, global datacenters accounted for 0.3 percent of carbon emissions in 2018, roughly equivalent to the entire country of Argentina.
[4]
Modeling something like carbon emissions from onboard AV computing can look simple, said paper lead author and MIT graduate student Soumya Sudhakar, "but each of those variables contains a lot of uncertainty because we are considering an emerging application that is not here yet."
[5]
[6]
There are unknowns aside from AV adoption to consider, and the MIT researchers admit as much in their paper. Level 4 and 5 autonomy, for example, "is still an active area of research and is proprietary for industry," the authors said, forcing them to use existing deep neural network self-driving algorithms as the basis of their compute power calculations.
Setting aside the potentially limiting assumptions made in the paper, the researchers say that in most of their scenarios, AV computing power doesn't just meet the aforementioned 2018 datacenter emissions number, but soars past it, unless onboard computers are kept to less than 1.2 kilowatts of power for computing.
[7]
AV compute power draw doesn't just have to stay below 1.2kW, either – the rate of computer hardware efficiency gain needs to pick up as well, from doubling every 2.8 years to doing the same in just 1.1 years "to keep [AV] emissions under [2018] levels."
"If we just keep the business-as-usual trends in decarbonization and the current rate of hardware efficiency improvements, it doesn't seem like it is going to be enough to constrain the emissions from computing onboard autonomous vehicles," Sudhakar said.
We're getting in on the ground floor
With autonomous vehicles still years from reality, the MIT research team reckon there's plenty of time to get out ahead of the problem of the carbon cost of AV computing, but it won't be easy.
Specialized hardware, for example, could reduce the complexity of onboard computer hardware and reduce its energy cost, but automobiles have a much larger lifespan than other types of computing hardware, meaning hardware could outlast software. That means AVs need to make a tradeoff between specialized and general computing hardware that, again, will increase carbon costs.
[8]Surely you can't be serious: Airbus close to landing fully automated passenger jets
[9]Tesla misses Q4 delivery expectations as stock keeps sliding
[10]Tired: Data scientists. Wired: Data artists
[11]City isn't keen on 5,000 erratic, traffic-jam-causing GM robo-cars on its streets
Algorithmic improvements could help efficiency too, but simplification could mean less safety, the researchers said, meaning yet another tradeoff has to be made.
That's not to mention the carbon use from burning energy to power the sensors and other parts of an AV, or emissions from AV construction, which the study said it didn't account for.
"This has the potential to become an enormous problem. But if we get ahead of it, we could design more efficient autonomous vehicles that have a smaller carbon footprint from the start," Sudhakar said. ®
Get our [12]Tech Resources
[1] https://www.theregister.com/2022/10/11/cars_not_self_driving/
[2] https://ieeexplore.ieee.org/document/9942310
[3] https://news.mit.edu/2023/autonomous-vehicles-carbon-emissions-0113
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Y8HijeZp0DIfHGWv69CtHAAAAAo&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y8HijeZp0DIfHGWv69CtHAAAAAo&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y8HijeZp0DIfHGWv69CtHAAAAAo&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y8HijeZp0DIfHGWv69CtHAAAAAo&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2023/01/13/airbus_future_technologies_subsidiary_wants/
[9] https://www.theregister.com/2023/01/03/tesla_q4_delivery_results/
[10] https://www.theregister.com/2022/11/08/data_artists/
[11] https://www.theregister.com/2022/09/27/gm_cruise_robocar_safety_waiver/
[12] https://whitepapers.theregister.com/
Re: Our potential future is a carbon-heavy one
Anyone who argues that there's a way in which a 1500+ Kg metal box that is used to move one person around is not detrimental to our climate, environment, safety, public finances really should come to terms with reality.
The power usage may very well go down
Although everything said is technically true, there is one key point that doesn't have to be: the power usage. They did the math with 840W - and I believe their numbers are right, when the computer uses it. But the self driving problem doesn't uses a infinite amount of processing power - sooner or later we will reach the "this is really good" point, and the computational power needed will not grow further. Then the tech evolution will take care of power usage.
So, it may use 840W today - but in 20 years time it may very well be down to just 20W.
Re: The power usage may very well go down
sooner or later we will reach the "this is really good" point
How many deaths per year will be the "this is really good" point where you think everyone will agree advancements can cease?
Setting aside the potentially limiting assumptions made in the paper: Good at being wrong
The numbers in the paper are a joke. The authors admit their assumptions are problematic, and that's an understatement.
Projections on computing technology in 2050 in made in the 2020's are going to fare worse than the ones made in the 1970s about the 2000's.
Ev's with self drive systems burning half the power of the drive system.
An exceedingly high vehicle to population number, and assuming most of those are sporting a car sized self drive using Kilowatts of power.
No provision for any kind of road adaptiations to make AV's job easier and more efficient.
So while 20w would be awesome, I'm not making guesswork for that far up the exponential curve of computing. I know that was a back of the napkin number and I'm not razzing you for it, just the authors. In car computing isn't passing datacenter computing. Any projections that don't account for the basic level of reality in those markets and don't line up with them are a huge red flag.
This paper and it's conclusions are sadly bullshit, but are one more piece of pseudo-knowledge that will be shopped around like it's gospel for a couple years, and then quietly and conveniently forgotten when it's projections promptly diverge from reality and people who put more work into the analysis than the headlines put out projections for a more reasonable time frame.
Rare earth minerals extraction (cost, environment)
If you are a realist like me, put in the account sheet the cost of rare earth minerals extraction into this, and you will see EV are WORST than fuel-based cars.
Re: Rare earth minerals extraction (cost, environment)
The scale of human suffering that's needed to make these electric cars is astounding.
That's a weird claim. A car has to carry around all the energy it intends to use. Currently, almost all of that is traction. If it ever gets to the point where a significant portion of a car's energy consumption is processing, we can be damn sure that engineers will try very, very hard to make it as efficient as possible - because it would impact mileage.
Or, I guess they mean car brains will run in datacenters? But that means that a car's ability to drive relies on an Internet connection. Nobody remotely sane would trust such a system with their life.
Exactly. The computer runs off the EV's battery.
I would think the sensors (RADAR, LIDAR, ultrasonics, cameras, pan/tilt/scan mechanisms, etc) would be more power hungry than the computer.
It will all depend how they get used
Its not just the potential computing power needed for self driving - a lot of that is happening anyway for additional reasons, but how usage affects overall mileage driven will have a big effect on the overall impact.
Many people point out that a personally owned car is very inefficient, and in one respect they are correct - my car is expensive in terms of resources and money to buy, but spends the vast majority of its time sitting idle outside my house or in a car park somewhere - a very inefficient use of an expensive resource.
But in another aspect my car is immensely efficient - virtually every journey it makes is a "useful" journey - taking me where I want, or need, to go. The only exception is once or twice a year when it has to make a short journey to a garage for new tyres or servicing - maybe 20 miles p.a. out of 20 thousand - that's 99.9% efficient.
One of the big arguments for self driving cars is replacing the personal ownership aspect with a more "efficient" shared style which makes better use of an expensive resource. But its worth noting that almost any form of shared usage, a traditional taxi, Uber, shared ownership whether traditional or driverless will result in a big increase in "useless" journeys - returning to base after dropping a passenger off, going to pick up a passenger, going between jobs. All this means extra mileage, extra fuel consumption (whether of the liquid dinosaur or busy electron variety) and extra road congestion. Now the more widespread the service the lower this useless journey overhead is, but it's never going to approach the 0.1% of the traditional ownership model.
Another aspect is if you have a personally owned vehicle capable of full service driving - how do you use it?. Obviously there are the standard usages - get work done during your commute to work, getting brought home legless from the pub. But there are other possibilities :- Don't want to pay more for your airport parking than the cost of your flight? - just send the car home again after dropping you off and then it can pick you up again on your return. Don't have parking handy near your city centre office or the restaurant you are going to - just send the car home - or maybe send it to find a free parking space out in the residential suburbs (very popular with the residents I'm sure). But this is all additional mileage, additional energy, additional congestion.
One other interesting question - how fast does a driverless car go when it's not in a hurry? - When I am driving or being driven I want to go as quickly as is commensurate with my comfort, safety and maintenance of my driving license - but if I'm say sending the car home after dropping me off somewhere I don't care if it takes 1 hour or 2 - whichever is cheaper in energy terms. We all know that 56 mph is a more fuel economic cruising speed on an motorway that 70 mph - how much better is 30mph? - that's going to cause fun!.
Re: It will all depend how they get used
> replacing the personal ownership aspect with a more "efficient" shared style which makes better use of an expensive resource.
Shared style usage might not be all that much more efficient than personal/private usage. It all depends on how things are being used. It's like with the trains. Lots of trains are rammed during the morning rush hour and the evening rush hour and very lightly used the rest of the time.
Add in all the extra road miles your post rightly predicts and we could be worse off.
As for your point of "let it drive home slowly" Oh boy is that one going to be popular, who's going to get stuck behind the slow moving energy save mode returning AVs.
High customer satisfaction
If they want this customer to be highly satisfied they could start by taking 90% of the computer crap out of the car.
Re: High customer satisfaction
Agreed. I can't find a single (US market) electric vehicle to even look at. That's because they all have satnav, embedded cellular and ten more pieces of sh*t that I don't want. It's ridiculous.
Go restomod
Plenty of classy old frames rolling around, if you aren't worried about crash ratings, or grab an old Volvo wagon if you are and aren't concered with looks.
Conversion will cost at least as much as the car, but imagine rolling around in a classic jag with all of the unreliable and fiddly bits removed.
And of course no need for the foil hats as the only modern electrics are ones you install.
On the other hand if the cars have good acceleration sense programmed into them they will save more fuel and have lower emissions than most of the human drivers on the road. Most human drivers seem to put little thought into what they are doing and are happy to leave their braking until the last minute and accelerate hard between junctions.
Also AI driven cars could communicate amongst themselves which would offer opportunities for efficiency that even the best efficient human driver would struggle to achieve.
Our potential future is a carbon-heavy one
We have a lot of carbon dominant items, cars, plastic bags, knives and forks, water bottles etc etc etc etc ... if we're going to still be here in a couple of thousand years then we need to change a lot of things, self-driving cars are just a small example of what we need to change.