August 23, 2015
Use Autonomous Car Computers When They Are Parked
Electric vehicle batteries will some day be used for load-balancing the electric power grid. While parked the EVs can feed electric power into a household or the grid while demand spikes. Machine learning systems will be able to predict future driving patterns and let a battery drain partially when the electric power is needed elsewhere. But picture the computing power in the parked autonomous EV. Seems like a waste when cars are parked for most of the day.
Each autonomous car will probably have more computing power than a tricked out desktop computer of today. Even fleet cars will be idle most of the day and night because outside of commuting hours most cars are parked. If the cars are EVs plugged into the wall they can just as easily also have their computer network plugged in. So then their CPUs could be made available for other uses.
To what use could we put hundreds of millions of networked autonomous vehicle computers aside from driving? The Folding@home protein folding distributed computing project comes to mind. Another protein folding project: Rosetta@home for disease research. Many other distributed computing projects for astronomy, astrophysics, molecular biology, mathematics, cryptography, and other fields of science could be done in networked car computers.
Update: One objection raised in comments is security. One way to handle that: Download code written in a virtual machine's opcodes (rather like Java) with no IO primitives in it. The code can do simulation calculations and have a way to say "I'm done" and nothing else.
Another objection raised: autonomous vehicles supposedly won't be parked much. But look at roads. How many cars are on them at midnight? Or at 6 AM? By contrast, how many cars on the road at 9 PM or 5 PM? There are huge spikes in percentage of vehicles moving. Travel times are very unevenly distributed.
Randall Parker, 2015 August 23 04:27 PM
Bad underlying assumption. If a car is capable of driving itself around, why should it stay parked 90% of the time. At this point, even car manufacturers are assuming that future of the industry involves far fewer privately owned vehicles and a lot of subscription to ride services.
I don't think so. Two problems:
1. Batteries typically have a limited number of charge/discharge cycles.
2. Automotive batteries will be optimized for automotive use.
So, what sensible person is going to let their expensive automotive battery's limited number of charge/discharge cycles get used up on load leveling?
If you don't pay for it, it's foolish for the car owner to agree. If you do pay for it, dedicated load leveling batteries designed for that purpose will be more cost effective.
The problem here, I think, is that you're subconsciously treating the automotive batteries as though they were some kind of super capacitor, with a purchase cost, but no degradation in use. Whereas the truth is that the degradation of the battery with each charge/recharge cycle is a major component of the cost of the car. It's kind of like proposing to stick alternators on IC cars, and auto-start and run them to use THEM for grid load leveling. It only seems to make sense if you ignore the costs to the owner.
"Load-levelling", if done without back-feeding to the grid, is simply demand-side management of the chargers. This does involve a longer time for charging, but if the vehicle is going to be parked anyway there is no inconvenience to the user. There should be no impact on battery life either.
AC Propulsion's V2G grid-regulation test actually increased the capacity of its (used, lead-acid) battery pack slightly. Small, short charge/discharge cycles appear to be good for some batteries.
The Utopian tone here is careless. First, recharging a battery EV requires so many KWh that it is comparable to running a central air unit for a whole home. If enough consumers have battery EVs, their recharging demand will quickly swamp the capacity of the finely balanced local distribution systems. It won't so much be the case that EVs can be used to "balance" demand, it will be the case that the charging adapter will be required to ask the utility if it can draw power, and charging cannot take place until permission is granted. Because of this, consumers will face the prospect that when they plug their car in after arriving home from a long day at the office, that their batteries might not be fully recharged by morning. Utilities must have this control of such large point sources of demand in order to protect the local system until its capacity can be boosted. And guess who gets to pay for rebuilding the grid?
A network with access to millions of vehicles? That would be a juicy target for terrorist hackers.
The most logical usage is mapping info that the car just collected on its drive, and uploading it back to the central mapping server. This will take a lot of capacity on its own.
"A network with access to millions of vehicles? That would be a juicy target for terrorist hackers."
Or the Chinese government, Russian crime syndicates and.... well the list is almost endless.
If it's a computer resource that I've paid for (or am making payments on) and it's connected to my wifi network, why should I not be able to use it for what I want? An autonomous car will have sensors and a lot of image processing built it. How about having the car spend its nights monitoring my security system, and even looking out for prowlers in my yard. If it's in the garage it could process video streams from any cameras I have stuck on the outside of the house. It could send a message to my smart phone or tablet (charging beside my bed) if it spots something suspicious.
If I use it for business it could track my milage, then upload the data to my business computer when it gets within wifi range.
Lots of other uses come to mind, other than ones that require opening it up to the big bad internet.
Hardware speculation: While a vehicle might have a lot of aggregate computing power, most of it will be partitioned into embedded or special-purpose computers devoted to a particular task; like a computer that controls a laser scanner and sends processed inputs on up to a Location & Mapping module. Devices like this are not easily re-purposed.
While EVs might spend most of their time standing around (by which I mean charging), the largest markets for autonomous vehicles will be long-haul trucking and urban transportation. Our current trucks could operate continuously if not for the sleep requirements of their human operators.
Urban transport shows daily usage patterns, but the city streets are not completely vacant after working hours. I suspect if public transport were more convenient and taxi-like, it would get more use outside the 9-to-5 window.
Speculative speculation: at night, a car will be busy "dreaming", processing the inputs it has seen to fine-tune it's neural networks and update it's maps.
This is about relatively small amounts of computer power connected over low latency, unreliable links. I doubt this is of much utility.
It would be pretty inexpensive for someone who wants to do this to just buy a bunch of used cell phones and hook them up over a fast WiFi. I doubt an automotive computer has any more useful capacity than a cell phone. You can buy used Android phones for $35 each.
recharging a battery EV requires so many KWh that it is comparable to running a central air unit for a whole home.
The average vehicle only travels 35 miles today. At 3kWh per mile, that's only 12kWhs per day, or 500 watts over the 23 hours they're not in use. And, of course, you'd want them to charge at 3AM when load is lowest (unless there's a peak in wind output), or at 1PM when solar peaks. For daytime charging you might need the autonomous vehicles to move over to the next neighborhood where power consumption is lower...