Driverless Cars

How Can Our Dumb Infrastructure Accommodate Smart Cars?

We’re going to need a lot more sensing equipment—and fast. Here’s how to do it.


By this point, driverless car technology has few skeptics. Videos of Tesla cars zipping around in full auto mode and Waymo's public passenger pilot in Arizona quickly deflate the idea that the software powering driverless cars is not viable. But we're not quite ready for a more autonomous transit system just yet. While some of our cars are quite smart indeed, much of our infrastructure is stuck in the Stone Age.

The problem is that driverless cars don't have a great way to communicate with the physical environment right now. After all, autonomous driving is not just a question of geonavigating a vehicle from point A to point B. It's a dynamic feat where the decisions a car must make will constantly change depending on things like the weather, the actions of other cars, and unexpected roadway intruders.

How to give cars better vision

For a while, policymakers focused on what's called "vehicle-to-vehicle" (V2V) communications systems as a way to give autonomous vehicles sharper vision. V2V tech focuses on intercar sensing, which would allow autonomous vehicles to detect and avoid each other.

Engineers had already been working on these technologies by the late-1990s, but in true government fashion, regulators put most of their eggs and funding behind a single technology standard, Dedicated Short Range Communications (DSRC), which was far from comprehensive enough to do the trick. After all, the armadillos that scurry across desert highways will not have receptors on their shells.

Today, engineers complement V2V technologies with a broader suite of what's called "vehicle-to-infrastructure" (V2I) transmission mechanisms, which would allow vehicles to also communicate with the physical world around them—through antennas, sensors, and cameras affixed to things like utility poles.

Policymakers are slowly but surely coming around to the understanding that emerging technologies will need more of this kind of "street furniture" to navigate our more connected world. Smart cars, drones, 5G connectivity, and "smart city" applications all need physical networking in order to properly function.

So the question becomes: how can we make sure that our dumb infrastructure gets smart enough in time for these technologies to roll out?

This is the problem that Brent Skorup and Korok Ray considered in their recent Mercatus Center study, called "Smart Cities, Dumb Infrastructure: Policy-Induced Competition in Vehicle-to-Infrastructure Systems."

State and local governments are going to have to get smart about how they incentivize street furniture installation and monetization. Skorup and Ray suggest a public-private hybrid. Municipalities should tap private companies to install the equipment, with one catch: open access must be baked in, so that different applications and technologies can benefit from the same infrastructure.

What do we need, and why can't we get it?

Skorup and Ray outline three broad categories of street furniture that we'll need: 1) basic infrastructure, which includes "passive" structures like utility poles, cabinets, wiring, and the rights-of-way necessary to install such structures; 2) network infrastructure, which is long-lasting networking equipment like fiber cables and data networks; and 3) the devices themselves, like cameras and roadside sensors.

To provide autonomous vehicle coverage across the nation, we'll need a lot of those things. A survey by the American Association of State Highway and Transportation Officials (AASHTO) projects that some 250,000 traffic signal locations (around 80 percent) should be V2I-enabled by 2040, and another 25,000 non-signal roadside locations will also be online. Then we'll need "accurate, real-time, localized traveler information" on around 90 percent of our roadways, and ideally more.

That's a lot of smart lights. And they will necessitate competent planning from the hundreds or thousands of nested municipalities that fund, oversee, and install such infrastructure across the US. Needless to say, this is a rather tall order.

A Government Accountability Office (GAO) report from 2015 dives into exactly what barriers stand in the way of such installing connectivity from sea to shining sea. Setting aside the spectrum management issues, GAO points out that many states and local agencies simply lack the knowledge and resources to adequately deploy and manage the enormous amount of smart infrastructure needed to power these more autonomous roadways.

Okay, so why not let private businesses take the lead? State and local governments already routinely contract the installation and maintenance of old-school telecommunications infrastructure to private companies. Couldn't Uber or GM or Google just duke it out amongst each other to provide infrastructure services for different locales?

Well, they probably could, but it would almost certainly introduce inefficiencies. The GAO notes that a key factor to smart infrastructure success will be "interoperability." Much of the street furniture that can fuel better smart car coverage could also serve 5G coverage, or drone communications, or even better garbage truck management.

Even if the networking equipment itself is incompatible, the rights-of-way that a company gets to install, say, a utility pole could also hoist up a 5G antenna. Granting a little networking fiefdom to a single private company could deprive a community of the benefits of open access, which means our environments would be much dumber than they needed to be.

Enter "policy-induced competition"

If policymakers proceed thoughtfully now, we can spare ourselves an artificially dumber infrastructure future. Skorup and Ray outline a model that they call "policy-induced competition" to overcome the dual problems of municipal limitations and private street furniture-hoarding.

Under policy-induced competition, governments would solicit bids for companies to install, maintain, and profit from certain kinds of smart infrastructure, just like they do for other kinds of networking equipment. But the governments would require winning firms to build access into certain categories of equipment.

Here's a good example of policy-induced competition: gas pumps. When you go to fill up your car, you don't need to worry about finding a "Honda nozzle" or a "GM gas pump" that will be compatible with your car. You just pull up to any gas station and know that the equipment will be compatible with your car. This didn't happen on its own, but was the result of a government requirement that all gas nozzles are the same size.

With street furniture, contracts to install or manage basic infrastructure (like utility poles) would come with requirements that the equipment is accessible to other companies and uses so that the community reaps the full benefits of these installations. The devices themselves—things like 5G equipment and cameras—don't lend themselves to interoperability mandates, and would be almost totally left in the hands of the private sector. Networking equipment is probably too complex for interoperability, too, but the authors leave open the possibility that policy-induced competition may make sense for this category of equipment in some circumstances.

Right now, state and local governments don't have much of a plan at all for how their municipalities will roll out the ample amounts of infrastructure and equipment necessary for not only smart cars, but also super-fast wireless coverage and smart city technologies. But the longer they wait to get their ducks in a row, the farther away the future of driverless cars will be.

Let's not waste our super-smart transit technologies on a dumb and unworkable infrastructure. If municipal planners get serious about implementing policy-induced competition now, our streets just might get smart enough to meet the challenges of our networked future in time.

NEXT: Trump Weaponizes the Bureaucracy Against Naturalized Citizens

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. So what’s the deaths and injuries per million miles for self-driving cars vs. human driven ones?

    1. I rather doubt that autonomous vehicles across all companies making them have a total of a million miles driving on real roads with regular traffic. Any number you could get at this point would be statistically meaningless.

      1. Way more than a million miles, and the accident rate is the same. But the fatality rate is much lower with autonomous cars. It takes a human to really screw up.

        1. Cite required.

        2. Oh, and the lower fatality rate is probably because the autonomous cars are mostly underweight econoboxes rather than the kinds of cars real people really want to drive.

          1. Uh, the tiny ones have the worst safety profile though. It’s why CAFE standards are directly responsible for hundreds of thousands of deaths.

            The better argument for the difference is that a much larger percentage of the autonomous miles are driven at lower speeds – most of the test beds are cities, and they do fewer highway miles. The worst overall safety record is Uber’s in Arizona, which had the highest average speed. Still though, the fact that they’re nearly at human parity already means it’s likely they’ll surpass human safety levels quite soon.

    2. hAnnAh. i cAn see whAt your sAying… elizAbeth`s storry is AmAzing… on sundAy i got A brAnd new hondA from eArning $9023 this lAst four weeks And even more thAn 10-k this pAst month. with-out Any doubt it’s the most finAnciAlly rewArding i’ve ever hAd. i stArted this 8-months Ago And pretty much strAight AwAy wAs bringing home over $71… per-hr. i use this greAt link, go to this site home tAb for more detAil…../


  3. This is a good article with good idea but what will be the cost to the public? Cities have a hard time maintaining infrastructure now and this additional infrastructure is likely to be even more costly. You can attempt to shift the cost to the private sector but I suspect one way or another it will come back to the public as taxes or rate hikes for services. Are we ready to take on these costs?

    1. I agree with the author that fully autonomous cars require smart roads. There are too many edge cases that cannot be addressed by an independent car, and inclement weather totally cannot be compensated totally.
      How will we get to smart roads, that depends on if you want to have huge programs to establish it, or let cities and towns compete for the best ideas with companies willing to use them for validation. I’m not suggesting experimentation on public roads but a joint effort that starts local instead of top-down. I see it as the only way to build up the knowledge before standardizing on a set of enterprise standards.

  4. As with electric cars, the poorer people in the country will pay a subsidy to the wealthy.

    1. just like with student loan forgiveness, disproportionately helping those with masters degrees.

  5. Don’t ya just love it when libertarian magazines publish articles recommending top down, government-driven central planning solutions ?

    1. I can handle reading things that I don’t agree with from time to time. Technological optimism is one of Reason’s big topics. And whether and to what extent government should build and maintain roads and other infrastructure is a big debate among libertarians.

      1. Not among libertarians. Among statists who can only think of top-down central planning.

        1. A lot of libertarians are also statists.

          This is also policy wonk shit about things that might actually happen rather than libertarian idealism. Which you seem to acknowledge below.

          1. Libertarianism is accepting of tiny and limited government as a necessary evil. Big difference between that and a statist who advocates massive government control over social and economic affairs.

            1. He’s simply wrong. The two are diametrically opposed and whatever he used to get to his conclusion should never be consulted again.

              1. How do you define “statist”? Most libertarians favor a state and a monopoly government. That’s a statist to me. Libertarians are just the lowest key statists with lots of limitations on the state, but that’s a difference in degree, not in kind.

                1. That is nonsensical babbling.

                  An anarchist believes in no government. We have a bunch of those here who claim to be libertarians.

                  Libertarians believe in a minimum necessary government to maintain common protection and in varying degrees, a minimum legal framework for the resolution of disputes based on torts.

                  Statists, believe in strong governments where rights are granted by states and responsibilities assigned from states.

            2. You think it is appropriate for the state to tightly control borders. How is that not statist? Just because the state you favor is limited by a restrictive constitution doesn’t make it any less of a state.
              Statists are diametrically opposed to anarchists. Libertarians are somewhere in the middle.

            3. Pragmatic libertarianism is accepting the fact that we are stuck with outdated government mechanisms due to history and inertia. And being willing to compromise in how quickly we do away with them, and what replaces them (if anything). But not being willing to compromise in the direction of newer or bigger government initiatives.

          2. “A lot of libertarians are also statists.”

            You mean like some shrimp are jumbo?

            1. Well, I think we need to agree on a definition of statist (or at least understand what each other means by it) or this is a pointless debate.

      2. So is abortion. It’s funny how Libertarianism is a big tent philosophy where it’s adherents proclaim that this batch of government money or this class of coercion is ok. I mean, it’s pretty much like calling yourself a libertarian doesn’t mean anything besides being a gun nut who doesn’t want rich people to pay taxes. I guess those last two things are sacrosanct so it’s good that we believe in at least something in this topsy-turvy world.

        1. Pay your mortgage and fuck off.

      3. Technological optimism is one of Reason’s big topics.

        Sure, but it’s rather orthogonal to anything resembling a libertarian topic. Unless…

        And whether and to what extent government should build and maintain roads and other infrastructure is a big debate among libertarians.

        No, it’s not. There are some utilitarians, optimistic futurists, and even statists that debate the issue, but the linking libertarian sentiment among them is largely ‘muh roadz’ and the fact that the while the government can and does, it isn’t needed or required to make roads happen.

        Herein lies the rub, as I mention further down the thread, discussion of alloying steel and building railroads isn’t anti-libertarian, but when you talk about current technology specifically to enable further government intervention policy-induced competition, you’re pretty decidedly in anti-libertarian territory.

        Being optimistic about the future isn’t anti-libertarian, optimistically dreaming up how the future government can induce competition in as-yet unfounded technological field through policy is exceedingly statist and anti-libertarian.

        1. Well, that’s how it looks to me.

          Of course, I’m sure your definition of libertarian is the absolute and undeniably correct one.

      4. V2I is not technological optimism.

    2. As much as this article’s slobbering over central planning disgusts me, I am not convinced it takes that side; it can be read as more of a review of current planning, not what the author or magazine) want.

    3. Luckily for us, reason is NOT a libertarian publication.

  6. I can’t wait to see an autonomous vehicle try to navigate my private, half-mile long, winding and frequently washed-out lane.

    1. Yeah, I’ll believe it when I see them doing my drive to work in a snow storm.

      1. to be fair, your neighbors aren’t so great at that either

        1. No, but it’s possible and most people manage not to crash most of the time.
          I do feel like the quality of winter driving has declined, though. And people seem to be getting more timid about it.

  7. Given the limitations on my own vehicle’s sensors in bad weather, I would like to see how a Tesla does in full auto mode in a snowstorm. Driving in perfect conditions does not prove viability in areas where those do not exist all the time.

    1. Global warming will take care of snow. See, its still all about the cars.

  8. This is nonsense. Central planning doesn’t work, and if the only way to get autonomous vehicles is nu cramming roadways full of 5G signals, there won’t be enough bandwidth or processing power available to do the trick.

    Humans get by pretty darned well with just one or two low-power eyeballs reading street signs and scanning for other vehicles, pedestrians, and armadillos. If computers can’t do that yet, then autonomous driving is not ready yet. Even if all that 5G infrastructure were in place, autonomous driving still requires vehicle awareness of pedestrians and armadillos, and that’s the hard part.

    Or to put it another way, recognizing and understanding street signs and signals lights is the easiest part of autonomous driving. Using 5G for that is a waste, and frankly just another excuse to spend my taxes on a bloated and expanding bureaucracy.

    1. How about we give an example. Self Driving cars were implemented perfectly in the first Grand Theft Auto game in 1997. Not just one car, but an entire city full of them, all following road instructions perfectly until the play comes and messes them up. However, that is because they are in the computer, the computer can perfectly know what every sign means.

      The problem with self-driving cars has always been sensing the road. Every attempt to fix that is either fiendishly difficult or an attempt to reinvent the train.

      1. Sensing the road is trivial compared to sensing pedestrians and armadillos. Autonomous cars have a long way to go.

        1. I can see you know very little, because road conditions are not trivial at all.

          1. I don’t think that’s what he claimed.

            1. It isn’t. I explicitly said “compared to”.

      2. No car is really ‘self-driving’. The difference is if the driver is built into the car.

        We can sense the road and there’s not really anything special about us in the senses department.

        1. Do you know how difficult it is to have a car tell the difference between a gust of leaves blowing in the wind and a wall?

          It’s even harder telling a pile of leaves or snow on the road from a big rock. Or a body.

    2. pretty much all of that. Reason’s obsession with self driving cars has always puzzled me.

    3. “just one or two low-power eyeballs”

      The bandwidth of vision is approximately equivalent to terabytes. Low power is wrong.

      1. don’t forget the direct electrical connection to one of the world’s most powerful quantum computers

        1. Your brain is not a quantum computer.

      2. Terabytes per what?

        Even if that’s true (I have no idea), the real power of human vision is that we don’t see everything. Our region of high resolution vision is actually very small. We are just really good at picking out the important things and ignoring the rest. A big problem with machine vision is that a camera sees everything the same. It’s just a bunch of pixels. Then someone has to figure out how to program something that makes sense of all that chaotic data. Our eyes and visual cortex just do all that without and conscious effort.

        1. It’s probably impossible to accurately guess what human vision would be equivalent to in computers. How much storage space is in the human brain? How many cute cat videos from YouTube can it store? No one fucking knows is the short answer.

        2. Its not.

          You can easily compute the bandwidth.

          Number of receptors (and this is easy as they’re all single color), number of intensity gradients, number of refreshes per second.

          About 9 megabits/s. Yeah – bits, not bytes.

          1. The more you know about human technical specs the more astonished you are that we can do more than drool.

    4. Why assume cars? Autonomous vehicles won’t be bound to just roads.

  9. This headline should have been the lead-in to a story about a VC firm attempting to buy roads from states or municipalities or voluntarily placing sensors to enable new tech systems. But it isn’t.

  10. our streets just might get smart enough to meet the challenges of our networked future in time

    for the next Carrington Event.

  11. What is the cost of smart roads versus smart drivers? I’m not talking about Johnny Cabs, I’m talking about Uber. Is the smart roads thing a solution in search of a problem?

    1. smart roads were proposed 30 years ago (at least). smart cars have passed them by.

  12. So Called ‘smart cars’ are yet one more techie idea that I think will be a disaster in real world. To err is human. To replicate that error ten thousand times a second requires a computer. I have dealt with buggy software some 56k was a lot of memory, and I have zero faith that smart cars will not turn out to be death traps.

    1. I’d agree if the current solution wasn’t to have average people driving all the cars. They are unquestionably worse.


      Nearly 1.25 million people die in road crashes each year, on average 3,287 deaths a day. An additional 20-50 million are injured or disabled.

      The number is about 40k dead in the US, and about 4.5 million “seriously injured”.

      Those are pretty easy numbers to beat, which is the true metric.

      Of course, nobody is going to use that metric. I’m quite certain we’ll go with “it killed a 2 year old on a tricycle!!!” as the metric, and the other 4.5 million folks who were not seriously injured or killed can go pound sand, because the blame for that is diffuse.

      1. I know people drive like idiots. I’m all too often one of them, though I try to keep in mind that I’m a poor driver at all times. What concerns me is the software bug that suddenly turns up in a thousand cars of the same model one Monday morning commute, when all of them suddenly turn left for no readily apparent reason.

        See, if the car is doing the driving, the authorities can say you should be paying attention all the time in case of error, but that just ain’t gonna happen.

      2. When you consider that the hundreds of millions of drivers drive trillions of miles every year in every imaginable condition, those numbers are going to be quite hard to beat. Driving is not a very dangerous thing to do.

        1. Yeah – we’ve made it very safe to drive mini-tanks around. Course it ain’t quite the same thing for pedestrians. We drove ‘pedestrians’ to near extinction in the US with ‘right turn on red’ which otherwise has great benefits. BUT – with American drivers it led effectively to intimidation of pedestrians off the road altogether. And with no ‘short walk for errands’ business anymore, the local biz just closed and had to go to the mall – reinforcing the end of the ‘short walk for errands’. Even with the numbers of pedestrians plummeting over those decades, the numbers of ped deaths/injuries did not fall at the same rate. And over the last 10 years or so with vehicle size still increasing and peds now virtually gone – ped deaths/injuries are rising.

          Driving is not dangerous. Walking is. Automating driving will not make walking safer.

          1. The annoying thing is that cars are the dominant form of transportation because of government policy at all levels. So much money is spent on roads; it’s ridiculous. City zoning discourages walking and biking. Parking minimums contribute to further sprawl.
            Federal lending policy encourages homes and businesses so spread out walking is about as useful as on the Mongolian steppe.

        2. Driving is the most dangerous thing you do. Except for visiting hospitals.

      3. Whenever someone says something is “unquestionable” then provides data that shows it is quite questionable i have to chuckle.

        1. I guess the science is settled. No point in debate anymore.

          BTW, what was the answer for when a self-driving car encounters a partially blocked lane? A human can go slightly left-of-center to get around the obstruction with no issue. No sane programmer would allow this, so the human has to take manual control, defeating the entire purpose of self-driving road.

          1. You don’t know a lot of programmers, do you?

    2. to be fair, dumb cars are death traps too. check the stats.

  13. This is a solved problem. And these companies already live in that space – tech companies cooperate on standards all the time. The web page you are looking at is the result of one of these cooperative efforts.

    If the government wants to help things along, they could probably set up some standards body or another, but the best thing would be to be as hands off as possible and let the industry players work it out, then support whatever they decide to do.

  14. “…Couldn’t Uber or GM or Google just duke it out amongst each other to provide infrastructure services for different locales?
    Well, they probably could, but it would almost certainly introduce inefficiencies…”

    Which is a very good indication that the entire enterprise should be ignored until the market says we need it.

    1. The competition among Publix, Kroger, Walmart, Food Lion, Ingles, and Aldi to sell me a loaf of bread creates certain inefficiencies as well. Bernie Sanders knows all about those inefficiencies and he knows how to eliminate them as well. Of course, some people might argue that eliminating those particular inefficiencies introduce other inefficiencies of their own, worse inefficiencies. But I suspect that if you’re starting your article with the assertion that there are few skeptics of driverless car technology, you’re probably not skeptical of Bernie’s fixes to the bread distribution problems, either.

    2. We all know what happens when the government makes 5 and 10 year plans.

  15. Couldn’t we just start working on something akin to a pneumatic tube system, like in Futurama At least my taxes would make going to work be fun.

  16. With street furniture, contracts to install or manage basic infrastructure (like utility poles) would come with requirements that the equipment is accessible to other companies and uses so that the community reaps the full benefits of these installations.

    Tom Johnson knew exactly how to get around that when he owned private street car companies. Just merge all the companies who acquire the contracts and hey presto – the private entity has their monopoly again. It’s why he became one of the best mayors in US history. He knew exactly how everything he wanted to do for the community could actually be undermined because he understood the actual value of municipal assets. Course that meant he didn’t sell those assets off for cheap either. The only way to ensure operational competition then was to maintain muni ownership of some asset that couldn’t be privatized and reassembled into a private monopoly. Course it was easy to demonize that approach as ‘municipal socialism’. We no longer in the US have the concept of ‘good government’. Nor do we much care about competition now either. Just eliminating it so our side can ensure a win.

    Notably missing in that V2V and V2I is V2P. Something as simple as ‘right turn on red’ pretty much eliminated drivers actually looking right before they turned right and thus eliminated pedestrians from being able to claim a de facto right to cross the road. Medoubts that Americans who program that V2 tech will start giving a shit about the ‘people’ part now either

    1. “…Something as simple as ‘right turn on red’ pretty much eliminated drivers actually looking right before they turned right and thus eliminated pedestrians from being able to claim a de facto right to cross the road…”

      1. I walk as often as I can. Even in a completely residential neighborhood, I’d guess that only 20% or so of drivers actually stop AT a stop line with enough intention/forewarning of that to be enough for a ped to take the risk and step out in front of the car anyway. The vast majority of those who stop only stop at the point where their vision to the left is good. And most (simple majority) drivers don’t even actually stop there either. They just slow down enough to be ABLE to stop and avoid a vehicle collision – but they are looking left for cars not anywhere else for peds – so the ped just has to wait, intimidated, off the road until the vehicle actually physically stops without creeping/encroaching.

        A couple blocks from my house is a rat-run type residential road. All the traffic is ‘going somewhere else’ and doesn’t stop/yield readily. With a public library and a senior/rec center. Not enough ped traffic for a light or a stop sign. But the yield sign wasn’t worth a shit – and those two facilities mean lots of kiddies and slow crossers. They put in a ped-triggered light – with IMMEDIATE light change. Literally as a ped you can hit the trigger and keep walking into the marked crosswalk as is the ped right. In reality, it doesn’t quite work that way since the first couple of cars almost always do what drivers do when they see a flashing yellow – accelerate.

        And the problem is getting worse because folks in SUV’s are absolutely the worst drivers for that. I suspect in large part because of the arrogance of ‘safety’. The same thing that happened with seat belts a bit earlier. Once drivers perceived they themselves were safer – they drove faster and thus made their vehicle more dangerous to others. So a 25 mph default limit on residential (15% fatality rate for an elderly ped if hit – which is already too accommodating to cars) is now a practical 35 mph (70% fatality rate for an elderly ped if hit) velocity.

        1. I’m familiar with Atlanta driving, “right turn on red” seems to be understood as “vehicles turning right on red have the right of way” because those bastards will pull right out in front of you while barely slowing down. I’m pretty sure they aren’t looking for pedestrians to their right when they’re not even bothering to look for cars to their left.

  17. “This didn’t happen on its own, but was the result of a government requirement that all gas nozzles are the same size.”

    That isn’t quite true. What happened was that EPA banned leaded gasoline and then set a maximize size for unleaded nozzles so that cars could be sold that would not accept leaded gasoline nozzles.

    Which federal regulation specifies the dimensions of a USB port?

    1. How is a cartel that much different than govt?

      1. Sarc or stupidity?
        A cartel can’t put you in jail.

        1. You don’t pay the USB cartel the fee they demand for compliance with their standard and guess who will throw you into a govt court even if your product complies anyway?

  18. By this point, driverless car technology has few skeptics. Videos of Tesla cars zipping around in full auto mode and Waymo’s public passenger pilot in Arizona quickly deflate the idea that the software powering driverless cars is not viable.

    Huh, apparently, someone hasn ‘t been paying attention to videos of Tesla cars careening around parking lots and decapitating their drivers.

  19. Because driverless cars are NOT working very well– which is why all of the industry players (except Tesla) have mostly abandoned all but level 5 driverless technology (realizing correctly that anything less is dangerous) the only way this will ever work is with a massive ground-based telemetry system, as I’ve commented repeatedly on these here threads. And that’s going to be very expensive, and a long time coming.

    1. A ground-based telemetry system also means that the cars would be useless outside of cities because nobody is going to invest in building that infrastructure out in Hicksville. You’re advocating for the “smart road” approach, but there are a lot of folks looking at this who say it is the wrong approach to go because we will never reach the point where the roads and guidance systems have the information needed. These people argue that you need to assume the roads and external guidance will be dumb and build systems that can deal with that in a fault-tolerant manner. The two approaches aren’t necessarily contradictory in that a smart car could use smart roads and telemetry when available and rely on its own sensors when that’s not there. But, if you design cars to *require* that infrastructure, they will fail when the infrastructure falls short, loses power, or has any fault.

      That tells me that relying on the ground-based telemetry approach is problematic compared to designing cars that can work in our current “dumb” infrastructure.

    2. I suspect it will ultimately be even more limited than that. Anything with a higher velocity than about 20 mph is ultimately constrained by how much we value a human life. Cuz that human is not automatable and the vehicle likely wins in any collision with said human. So any public land with a higher velocity than that will de facto eliminate humans from access.

      Of course, I suppose its possible that said robocar could be configured with a giant spike coming out of the steering wheel with seat belts strapping the passenger to those spikes. Or maybe the only seats are exposed right on the front bumper so the vehicle owner actually becomes the most likely victim in a crash. Methinks that wouldn’t sell too well. But other than that scenario, all the ‘safety’ improvements would be designed to improve the safety of that owner at the direct expense of every other human.

  20. If the driverless cars are so much smarter than human drivers, why do they need $billions worth of technology that humans get along without?

    1. You have way more than a billion dollars of technology in your little finger.

      You are, literally, a nanotech machine.

  21. Or, we could just forget this stupid idea.

    1. It is a very niche thing that’ll only work in only a few areas for the foreseeable future.

  22. After all, autonomous driving is not just a question of geonavigating a vehicle from point A to point B. It’s a dynamic feat where the decisions a car must make will constantly change depending on things like the weather, the actions of other cars, and unexpected roadway intruders.

    Actually, there are more problems geonavigating a vehicle from point A to point B then just the dynamic factors no on a map.

    GPS precision and what level of GPS is being used in the vehicle.

    Consumer grade GPS is inexpensive, but precision can vary with conditions, including season, weather, land topography and proximity to large artificial structures*. Precision ranges from within 4 meters to within 32 meters,enough to actually miss a road, and even 4 meters isn’t sufficient to hold lane position.

    Map grade GPS can get you to within 1 meter, but the receivers cost upwards of $1,000 each.

    Survey grade GPS can get you to sub centimeter precision, but the receivers cost upwards of $10,000 each.

    Then there’s the map data itself, which is never 100% accurate. There are documented instances of crashes or people driving off the ends of roads into bodies of water because a human driver was paying more attention to their GPS navigation system than to their surroundings.

    The idea that an autonomous vehicle could rely on geonavigating from GPS data just to stay on the road without using local sensor data like cameras is laughable.

    1. *Large metal and/or glass structures can reflect GPS signals making the receiver think it’s farther from a given satellite than it really is, throwing of the triangulation calculations used to determine the units position. Higher grades of GPS use more satellites at once so they aren’t as affected by this.

    2. GPS won’t cut it. It needs machine vision and in the end, a massive ground-based telemetry system so the environment can communicate with the car. This will be a combination of sensors by the road, under the road (in some cases– complex intersections, urban environments) and two-way communications telling the car what the condition of the traffic and roadway is etc.

      Self-driving technology will happen, but it’s a long, long way out. The idea we have all the technology we need to stuff into the car and only the car to be able to navigate the world– that’s laughable.

      1. The idea we have all the technology we need to stuff into the car and only the car to be able to navigate the world– that’s laughable.

        Kinda why I wish Reason were a little more pragmatic or honest in it’s science reporting or futurism. Borrowing from Heinlein*; I’m fine with discussions on railroads, but the confusion induced by discussing how to alloy steel so that the government can construct railroads for everyone’s benefit is exceedingly irritating from a libertarian standpoint. The fact that, depending on the author, I’m being generous assuming the confusion is being accidentally induced makes it all the more aggravating.

        *”When railroading time comes you can railroad—but not before.”

      2. That was my point. But the article mentioned geonavigation which is very specifically based on the combination of GPS and electronic map data.

    3. We have a word for geonavigating. Its called navigating.

      1. geonavigation is something specific. Specifically navigation through the combination of Geo-locating technology like GPS with electronic map data.

  23. The biggest impediment would seem to be lawyers.
    One wonders how it only took about 35 years to develop a national paved road system for the early autos.

  24. Is anyone else getting the feeling that the whole smart cars vs. smart roads debate has been a huge bait and switch?

    For the longest time smart roads were going to be too expensive and weren’t going to work for most of our infrastructure, so smart cars were the ‘only’ option. Now, it seems like smart cars won’t entirely fit the bill and we’re going to need at least some smart roads as well.

    Smart cars weren’t the solution to whatever problem they didn’t solve, they were the solution to the problem of getting taxpayers to foot the bill for more surveillance and smart road technology.

    1. “Now, it seems like smart cars won’t entirely fit the bill and we’re going to need at least some smart roads as well.”

      They won’t even come close.

      1. They won’t even come close.

        More to my point, it was a false choice. The bullshit optimism about freedom and cost reductions associated with self-driving cars was more of a lie to get just the tip and make way for the shaft of more controlling and expensive infrastructure.

  25. “By this point, driverless car technology has few skeptics.”

    Narrator “she’s drunk”

  26. How about we build cities around people first huh? Seems a far smarter idea than building it around yet more cars.

    1. I’m pretty sure that will have to happen in every other country instead. Car marketing in the US is the most successful ever. Most American men would rather castrate themselves than get a smaller car. Indeed even the red sports car as a sign of middle age has all but disappeared now since the visibility now is horrible and its just not safe to drive that size car. And the American woman would rather sell her kids to a pedophile than reduce the amount of cage around her precious’.

    2. How about we build cities around people first huh? Seems a far smarter idea than building it around yet more cars.

      Unless you’re talking about tearing out/down the cities that exist, it can’t be done. Building cities around people first means you can’t build them any larger or higher than average people can walk in a relatively small fraction of a day and even relatively small cities are too large for such a policy. You could work various public commuter options in to expedite the commute and enhance the density but then, of course, large parts of the city would be designed around the commuting systems rather than people.

      1. You CAN build neighborhoods around people. Which means dramatically rethinking the grid system. Once you have people-friendly neighborhoods where the grids aren’t just slicing/dicing them up, THEN you can think about how to connect neighborhoods. And hey presto – NOW you have people-friendly cities.

        It’s interesting that ‘the suburb’ – designed entirely for the car – also got rid of the grid system because that grid system is simply death for controlling the car absent physical barrier stuff like bollards/etc. So suburbs have their windy dead-ends and cul-de-sacs and limited entry to arterials and such. Which is technically good for walking/biking/etc away from those arterials – and keeps the car velocity up on the arterials. Problem for suburbs is that they are mass-zoned so that there is nowhere to walk or bike to that’s within many miles. And they can’t fix that – grids can.

  27. “By this point, driverless car technology has few skeptics.”

    OK, Zoomer.

  28. In all this discussion of driverless cars, I’m shocked nobody has brought up the Red Barchetta problem.

    1. Yep. A “driverless” car will take you only where the programmers have decided you’re allowed to go. Or nowhere at all, if that’s their decision.

      1. Or anywhere. That’s another possibility.

        I agree that driverless cars are a huge opportunity to control people, but that’s not the only possibility.

    2. And all this talk of having the cars rely on outside information, and the term “spoof” appears nowhere.

      Sure, if all the cars are communicating you can do cool stuff like having all the cars accelerate at the same time when the light changes, instead of each waiting for the one in front to start moving. But then some idiot decides it’s fun to spoof the system, and you start getting rear ended.

      You know it’s going to happen, if the cars don’t just go by what they can see.

  29. No no no no no government at all. 100% privately owned censor/camera cooperation within big tech while “leasing” the smart infrastructure sites from the locals, states, feds…think forest land leases for ski resorts. Stop sacking the taxpayer and writing prog planning bullshit!

  30. I’ve seen plenty of street furniture already. Usually it’s too old and beat up to bother taking for my apartment.

  31. My prediction: smart cars and roads will happen soon, and first on selected routes in urban areas and commuter paths. That will reinforce the special infrastructure needs, from clear pavement markings to high tech sensors and com networks. It will also demonstrate the challenges that non-smart, i.e. human drivers, pose to both infrastructure and to the smart cars. And that will lead to bans of human drivers on the smart routes.

    1. Which will accelerate our descent into civil war and schism.

  32. I think the thing that amuses me most about this issue is that ‘self driving cars’ are one of those things that will never work but here we are asking ourselves if the same people that programmed Boeing flight navigation should be in charge of every vehicle in the country in far more unpredictable and varied conditions.

    Yeah, sure. It doesn’t even matter if the computer can have fewer fatal accidents than humans. What matters is that you can’t cause 300 million brains to all fuck up at the same time shutting down the entire nations economy in about ten seconds time. It doesn’t even need to be a bad actor, it can just be incompetence and a bad software update. Or a busted sensor from road wear and tear shutting down an entire city district.

    That any moron thinks otherwise is a testament to idiocy and shows how much faith people have in systems designed by fallible and frequently wrong other humans. They think of computers as gods of our own making, rather than tools that not only share our own flaws but who have entirely new flaws we invented for them.

    1. On cost alone you’d never in a million years be able to upgrade the entire United States infrastructure to support so-called ‘smart’ cars. We aren’t Europe. We aren’t France. It would take decades to upgrade our entire road system, and by the time you were done it would all be outdated technology and you’d already be halfway through the next upgrade cycle. Oh, and it’ll cost at least double what it costs now unless we’re talking about using solely Chinese or Indian computer systems in our national transport network.

      What could possibly go wrong?

  33. Good god, this article is off base.
    The fact that autonomous vehicles need tons of additional infrastructure and can’t navigate using the existing driving information infrastructure (lane dividers, cross walks, bike lanes, stop signs, yield signs, speed limit signs, stop lights, turn signals, etc…) safely and effectively is proof that they aren’t anywhere near ready for prime time. If an invention needs infrastructure and others to accommodate it’s short comings, it’s not worthwhile or needs further development.
    And the gas pump nozzle standard example is a silly example of the need for government mandated standards. This web page and the computer or mobile device you are viewing it on uses hundreds if not thousands of open standards developed by the free market (whether it’s individual companies or voluntary consortiums) and none were mandated by the government.

    1. This web page and the computer or mobile device you are viewing it on uses hundreds if not thousands of open standards developed by the free market (whether it’s individual companies or voluntary consortiums) and none were mandated by the government.

      Notably, those systems can’t crush you dead or drive through the lobby of a department store at 90MPH either. You see, we tend to think of systems that can directly murder us differently than systems that provide us with communications or cute cat videos.

      Not that I necessarily disagree with the sentiment of open source and/or industry guidelines, but I stop thinking that’s a good idea when a system can actually directly murder people.

      1. And you think government designed / mandated standards for systems that can kill people would be and improvement?


      2. I disagree with your assessment about they should be treated differently. Here’s a direct counterexample that works just fine and is in circulation now about voluntarily non government adopted standards in safety critical applications.
        I’m a glider pilot. There is no regulation about standard locations or color coding of certain controls but the industry have pretty much converged on certain flight controls being in a certain cockpit location and a certain color. And BTW, there have been documented cases of accidents caused by confusion of which flight control is being manipulated with uncommon cockpit layouts in unfamiliar aircraft. The industry adopted a de facto, informal but not legally enforced standard, how about that.

        1. A less serious example is motorcycle controls.

  34. Here’s a good example of policy-induced competition: gas pumps.

    I’ll see your gas pump fallacy and raise you a spill-proof gas can.

  35. This article’s premise is seriously undercut by the fact that gasoline nozzle sizes were standardized well before the government regulation. The government only got involved when they decided to 1) mandate unleaded gas, then 2) had to figure out how to stop people from continuing to put leaded gas into their unleaded cars. (For those who don’t remember, your unleaded engine will run just fine on leaded gas. However, the lead destroys the catalytic converter – another government mandate pushed on sometimes-unwilling consumers.)

    Government “solving” a problem that it originally created is not an example of “policy-induced competition”.

  36. I look forward to the day when Google or Twitter or Amazon is driving your autoautomobile and when they detect you saying something they disapprove of the ejector seat dumps you on the curb.

  37. I’ve got a better idea about self-piloting cars on public roads. Make the ones responsible for their programming criminally and civilly liable for any harm done to everyone outside of them. Wringful death cause the stupid smart car did not sense th victim? Criminal negligence. Charge as if the assault/death was wilfully caused by the one in charge of the car.

    1. I still prefer – the only seats are strapped to the front bumper. You crash – you die. Saves money on the lawsuits and the govt courts too.

  38. The problem is that driverless cars don’t have a great way to communicate with the physical environment right now.

    Neither do you.

    This isn’t going to be a pro-V2I article is it?

    For a while, policymakers focused on what’s called “vehicle-to-vehicle” (V2V)

    Yes, but V2I was actually what was focused on *first*. Somewhat later V2V was proposed as an alternative.

    So, yes it is a pro-V2I article – in a libertarian magazine. We’re actually now talking about how good an ubiquitous surveillance system would be to install across the country?

    And I remind you that YOU are perfectly capable of driving from LA to NY without a single but of active infrastructure. And its becoming increasing obvious that a vehicle could do so also. So, no, let’s not spend billions of dollars on fixed infrastructure that will be either – obsolete before the first bit is installed (government control) or utterly incompatible from state to state (‘state level’ government control). Because we’re not going to pretend for an instance that any of this would be private or that competition would be allowed in.

    What’s next, Reason’s case for light rail?

  39. What a surprisingly statist viewpoint to find in Reason. V2V and V2I are not something any real robocar team actually plans to use. Their utility is very low, their cost very high, their chance of working very low, and their addition to computer security risk high. They never really developed so the FCC is proposing to return the band to the people by making it unlienced.

    We want smart cars and dumb roads. That’s what made the internet succeed — infrastructure that is as simple and stupid as possible, and all the smarts in the devices, not the network. Cars (and even more, phones) are replaced frequently, infrastructure takes decades to plan and lasts for decades — it is completely at odds with computerization.

    For more details see my essay on this:

  40. After reading “Passengers”, smart cars are a dumb idea. I’ll gladly take my chances driving my own vehicle (not an EV) my own way. Computer programs always have glitches and are easily hack-able. I don’t care to be killed by some stupid Microsoft program that goes haywire. It isn’t about saving lives, it’s about selling technology that will cost the consumer a boatload (it already does). And who will pay for smart roads and the infrastructure? The passengers will, not the computers driving the vehicles.

  41. If the driverless car’s software is not free/libre, that means some company can design that software do whatever it wishes, and the users can only take it or leave it.

    Nonfree software nowadays is generally full of malicious features (we’ve documented hundreds of examples in, and cars will be no exception. In fact, we know that the software in the Tesla records where the car has gone. This would be ideal for the Chinese regime, or a future American regime.

    But it gets worse for a driverless car, because its software could be designed with a back door that allows someone to send remote commands to direct the car to take you to any specified place — perhaps the nearest reeducation camp, or CIA black site, or Homan Square. Private parties (criminals) might also figure out how to use the back door.

    The only known way to make sure the car can’t do this is if the users have control over that software. That means, free/libre software. See

    With free/libre software, the users have control over the program, but separately and collectively, so they can make sure there are no back doors or surveillance malfeatures in the programs in the car.

    But that remedy is usable only if the car is your property. A taxi in which you are a passenger is the property of someone else; the owner would be allowed to modify the free/libre programs that run the taxi, just as the owner is allowed to work on the engine, wheels and brakes, but not the passengers who only ride in it. So how could you trust the driverless taxi?

    The only way you can trust it is if it has no way of knowing who the passenger is.


Please to post comments

Comments are closed.