A couple weeks ago I had the pleasure of riding in a friend’s new Tesla Model X. It parks itself effortlessly, and as my friend engaged the autopilot and let go of the wheel at 70mph, my trepidation quickly evolved into a sense of awe and wonder. For two hours from Mt. Vernon to Seattle we barely touched the wheel. It recognized speed limits and adjusted instantly, maintained safe following distance, and on neighborhood arterials even recognized the difference between cars, buses, and cyclists. It was incredible. There was a undeniable giddiness at the sheer elegance of it all.

As emotions go, wonder gets its hooks into us pretty easily. By exceeding the limits of our knowledge, opening up new paradigms, and projecting authority, wonder tends to provoke an almost religious response. Specifically, wonder generates trust and faith.

This trust and faith too easily becomes salvific. We pledge our allegiance to the hammer, and suddenly everything looks like a nail. The technology is so good at performing its tasks that we forget to ask if it’s performing the right tasks in the first place. We conform our lives to it, rather than using it judiciously for the tool it is.

There’s a temptation to think that solutions are really complicated, and that we need creativity and genius to solve them. There are things like civic hackathons and large-scale competitions such as Challenge Seattle that do really good work at optimizing what we have, but they also share the conceit that if we were only a little smarter we could figure this all out.

At its best this impulse drives genuine innovation that makes our lives richer and more connected. At its worst it is a hubris that causes us to overthink and overengineer the solutions to our problems. But as our mortality attests, we are still small bodies in a large world, with bodily geometries and needs no different than our ancestors. To me, the striking feature about our daily lives isn’t their complexity but their banality. Employing neural nets to get a loaf of bread is excessive complexity to say the least. Relying on them to systematize all of our travel seems crazier still.

This isn’t to be a Luddite, or to claim any sort of purity from eschewing or scoffing at technology. It’s simply to not ask those technologies to do unnecessary or wasteful things, or to seek from them miraculous relief from basic geometric facts.

I see driverless cars as a necessary but by no means sufficient transportation innovation. They will help us wring out more road capacity, provide good arguments against road expansion, possibly enable less urban parking, and they may yet revolutionize taxi or bus-based transit networks. But there will still be fundamental geometric limits to these improvements. Any reliance on them in urban settings will disappoint, bumping up against the familiar feeling of too many vehicles in too little space. And to the extent that they tempt us to divest from mass transit, walking, and biking, they will actually worsen the status quo.

But what if instead of attempting to fit more of the same sand into the same hourglass, we just shrunk the grains of sand? Like the classic GIF so tautologically suggests, when you strip away the requirement that humans be encased in metal (autonomous or otherwise), you suddenly see not a scarcity of space and a surplus of people, but instead a scarcity of people and a surplus of space.

The Occamite solution is perhaps the most elegant of all, and it thinks more like Venice than Silicon Valley. It doesn’t make the reflexive mistake of nostalgia for an earlier time,  when air and water were dirtier and cities were often places of dense misery and disease. Instead, it enables us to integrate technology toward a healthier and richer human experience, but one that’s still human-scaled. People scoff at the the “19th century technology” of trains, but you never hear the same disdain for the cellular respiration that moves our feet or turns our bicycles’ cranks. It’s ok to be small.

The Tesla I rode in and drove was magical, but the following morning it wouldn’t start until it completed a software update.

75 Replies to “The Overengineering Temptation: Or How Driverless Cars Solve the Wrong Problem”

  1. Love the idea of being driven about in a computer with seats and wheels, but putting my faith in the kids whose parents are walking and cycling and Link’ing everywhere with them.

  2. They will help us wring out more road capacity

    How’s that?

    A computer still needs adequate space to stop in the event the car in front suddenly gets a flat tire or some such. The stopping distance is governed a little by reaction time, but more by sheer physics.

    Therefore, I really don’t see that much, if any, road capacity improvements being made. People are already driving far in excess of safe following distance (recommended 2 seconds, but actual stopping distance is typically four seconds) because they don’t care if they kill themselves or someone else. Computer drivers, however, will have been programmed to maintain a safe distance.

    1. Simple – Cars talking to cars can decide who is going where, and ‘Platoon’ themselves together with very short headways, while increasing safety because of all the real time sensors. Distracted drivers, with limited reaction time is a big problem in accidents.

      1. Right, in theory that is how it is supposed to work. But until then, it will probably be worse. I would assume that the cars are programmed to travel at a safe distance. But drivers routinely take chances, and assume the best. We are willing to accept some risk in the hopes of getting there faster, but I would be very surprised if an automated car would be programmed to do that. Even with top quality gear, by the time the car detects a slowdown ahead, it would be too late. So once everyone is riding an set of networked cars, then platooning can occur. But until then, automated cars won’t be able to wring out more capacity.

      2. This works for merging and turning, but it doesn’t work for assorted other issues such as flat tires, sofas that may come flying off of pickup trucks or various other problems that require stopping distance.

      3. Killer (literally) problems, Mic and everybody else, because Zach already knows.

        1. A computer only knows the last thing the dumbest, worst character-flawed, sadistically malevolent, ISIS-employed, brain-diseased, talk-radio-warped, Twitter-addicted human told it after his last snort of his worst meth just before his works blew up and melted him.

        2. But even so: If he’s driving the bus in person, at some level of his damaged mechanism, he gives a crap whether he gets killed or not. And from a million years of sweet lady monkeys before his family burned off all their tails at family firm start-up, part of his DNA cares how many children he kills.

        3. And between him and a computer program- which one wouldn’t need a drug test to not get hired?

        Cue theme from “Deliverance!” Great posting, Zach.


      4. Autopilot cars are paranoid about safe driving distance now because they’re still gaining acceptance. “My trepidation quickly evolved into a sense of awe and wonder.” In the future greater public acceptance, technology advances, well-tested liability law, and a lower percentage of human-driven cars may allow the air bubble to be shrunk. But “Cars talking to cars can decide who is going where” is like telephones: they only work for you if all your neighbors have one, and if they speak a common protocol. In the early days of telepony, an office sometimes had multiple telephones for different local networks. These automakers are each designing their own technology for first-mover advantage. Do they have a common communication protocol? Not just the physical transmission but the messages. Or have they not gotten that far yet? Are they cooperating or is each one trying to bully the others into accepting its standard so it gets billions in licencing fees?

      5. You can see my comment below about the idea of cars driving like cabbies ( I doubt it will ever happen. They will drive like driving instructors. If there is a “New York cabbie” setting, I doubt it will be available to regular people (it will be available for emergency vehicles). How would that work, anyway? You tell your car you are willing to accept some risk, just so you can get there a couple minutes early? That will never fly. How about speeding? That is quite common on the freeway, but I doubt that will be acceptable with the software (cops would never allow it). Which brings up an interesting subject — I wonder what happens when some racist cop pulls over yet another black man even though he is driving the speed limit. Good God, I want a part of that lawsuit.

        If there is an improvement in speed, it will be for the group as a whole, not the individual. Not just when all the cars are automated (a huge jump ahead — there were plenty of horses on the streets for quite some time) but in the short run. I could see it working like ramp meters. Your car simply won’t allow you to merge, because it is for the greater good. If you are behind that car (and driving) there is nothing you can do about it. It certainly won’t allow you to change lanes willy-nilly, hoping that the other lane is faster. This means that a driven car will be faster, because a driven car can be a lot more selfish. At some point those driven cars will become such a small percentage that it doesn’t matter. They will be like Model T cars, slowing down the traffic on the highway. Very rare, but perfectly legal. The weird part is that unlike the model T, they might actually be moving faster.

      6. “You tell your car you are willing to accept some risk, just so you can get there a couple minutes early? That will never fly.”

        Especially with the reversal of liability to the car manufacturer. You can drive risky or break the law when it’s your own liability, but when it’s the car manufacturer’s liability and you have no way to assume it, the cars won’t let you do it. Even if a personally-owned car has a waiver option (“I accept liability”), a taxi won’t. Hmm, that could create ongoing demand for old-fashioned cars, especially in more conservative/libertarian areas, to preserve your constitutional freedoms (and ability to speed away from a chasing cop).

      7. Platooning is unsafe, illegal, and a liability nightmare. Will Never Happen. Pure fantasy material. Automated cars will keep a safe following distance…. or the *programmers will be liable for wrongful death suits, in the hundreds of millions of dollars*. You know what they’ll do.

    2. I think the quote should have been “allow us to better utilize the existing capacity we have”. With less slowdowns and gridlock, due to bad driving, we’ll be able to maintain closer to the theoretical 2000 pcphpl that most freeways are designed to, even during rush hours.

    3. The big capacity gains will be on closed access, high speed roads. I think RossB and Glenn’s criticism will be more of an issue on urban roads & side streets.

      1. It is still an issue, though. Imagine a bunch of cars driving on the freeway, with a half second between them (fairly common, although obviously unsafe). A semi blows a tire. What is a driver (computer or human) supposed to do? Slam on the breaks, switch lanes? No matter the choice, it will likely be a mess. That is why I think until the whole network is connected, and all those vehicles can slow down in unison, the software will choose a more safe distance. That has been the case so far. People have said that the Google cars aren’t that aggressive. They are too slow. They are like the driving instructor, not the New York cabbie.

        Until these become very common, every report of an accident, no matter how small, becomes news, and people selling the technology hate that. There is bias against technology. We still have pilots, even though a lot of studies suggest we would be better off without them. But we still want pilots to control things in case “things go wrong”, even though it is way more likely that a pilot will be the cause of the problem, not the solution.

    4. Glenn, everyone knows computer driven vehicles are not governed by the laws of physics!


      A single driver can entirely eliminate a traffic jam behind him, simply by leaving more space in front and minimizing acceleration and braking.

      I suspect there is some definable tipping point function that gives a minimum percentage of knowledgeable and proactive drivers optimizing traffic for throughput instead of personal advantage to entirely eliminate traffic jams.

      With an increasing percentage of driverless cars on the road, I expect we will land on the free flow side of this function quite soon.

      1. Funny, I just posted much the same thing. I personally think it will take longer to reach that point (a few aggressive drivers can also cause a traffic jam) but I think the conclusion is the same. There is a very good chance that — like ramp meters — individuals will be slowed, but the collective group will see better throughput. Of course, that might mean waiting a long time at on-ramp. Buses and trains have the potential of moving people much faster, as they simply take up less space while doing so.

    If you haven’t watched Tony Seba’s talk to the Seattle Downtown Assn earlier this year, it’s worth 17 minutes of your life to glimpse into the not so distant future.
    He starts with a photo of NY’s 5th Ave in 1900 where horse drawn carts filled the streets. Just 13 years later, those carts are now cars, with only one horse to be found. Rapid advances due to technology doesn’t take long to catch on.
    Tony follows with several more examples where the pundits got it all wrong, such as cell phones and computers. Driverless cars talking to each other and ourselves will likely be commonplace by the time ST3 is built out. The implications are significant. Watch the video given to our movers and shakers, then decide.

    1. Do you have a transcript for his talk (or a similar article)? I prefer reading over a lecture.

      1. Not from the talk itself, but you could get his book online.
        From Parking to Parks: Cities and Self Driving Car Disruptions
        Tony Seba is a Stanford Univ Professor, futurist and worldwide lecturer on energy and transportation.
        He asserts that by 2030 we will have far too many roads, car ownership will be the exception, and most of Seattle’s 100,000 parking spaces downtown will be surplus.

      2. Thanks mic. I will try and find a shortened version of his book (maybe a magazine article). (I know — I am picky — I want to read, just not a whole book).

    2. Streets filled with horse carts

      13 years later…

      Streets filled with cars

      Note what hasn’t changed, not until we figure out teleportation.

      1. There’s a longstanding meme that the width of street lanes and railroad tracks is based on Tim Eyman, I mean a horse’s ass [1]. Because the form factor of trains and cars is based on the form factor of carriages and chariots.

        [1] Initiative 831 officially declared Eyman to be a horse’s ass.

      2. Tim Eyesoreman is more than a horse’s ass but also a coward who lacks the guts to charge the hill. He’s also a campaign finance cheat and hates transit because he just hates the idea of having to share a ride.

      3. Horse’s asses, actually.

        This width of 56.5 inches was set by an act of British parliament in the Regulating the Gauge of Railwsys Act of 1846.

        So, horses ass isn’t too far off. It’s just the vast quantity that the meme gets wrong.

    3. I’d second that Tony Seba/Seattle Downtown Assn link. I watched it a while ago and it was well worth the time. I doubt that a good future will involve widespread ownership of self-driving cars, but vastly reducing parking and personal car ownership in urban areas is a quick win now.

      Also, BRT will be serving more routes in the near future with wider stop spacing and wider route spacing. I’m interested to see if small shuttle routes (possibly with on-demand booking or DART-style routes) could use a self-driving system similar to the CityMobil2 routes that are already deployed.

      Self-driving technology absolutely isn’t is a good argument against further investment in high-quality transit.

      1. I would agree Tim. The implications of smart cars is this. You don’t need mega parking garages to store the days catch of commuters. Today, most cars are only on the road 4% of the time. The rest is being parked. Shared vehicles could just keep feeding transit stations and rarely stop, or at least not at the transit center.
        Curb lanes now dedicated to parking now become transit lanes, and lots of them for cheap.
        Neighborhood shuttles become self driving vans, not DARTs.
        And on and on.

      2. Small shuttle routes is what Uber wants to evolve into (among other things). I can see transit agency’s contracting out to Uber like private providers for DART like services. Would be especially useful to feed suburban neighborhoods to LRT and BRT stations. I can imagine a future where a commuter in a subdivision pays a subscription and has a driverless van pick her up at her front door at a set time and delivers her, along with a handful of her neighbors, to the closest transit station. Coming home, there will be a queue much like a taxi queue at an airport, except she’ll share the van with a few other people and the van will know her address when she gets it.
        This driverless van will be busy shuttling people to and from the transit center 2~3 each morning & evening during peak commute, and otherwise be on call for off peak commuters.

        I wonder how driverless cars will change the calculus around Park & Rides, both in terms of people getting dropped off rather than driving themselves, and driverless cars being able to park more compactly?

      3. RE: Shared cars – the Car2Go cars already sorta show this pattern, with many cars moving to the residential neighborhoods at night, spending the night there when there is little use, and then driving back into job centers during the day, where they may be used multiple times until they head back “home” for the night. So there may still be a “day’s catch of commuters”, as mic puts it, if people still plan on driving into the city each morning. Reverse commuting out of Seattle offsets this, but not nearly enough to prevent a mass of cars parking downtown each day, driverless or not. One of the advantages of a traditional transit system is after those packed trains/buses drop people off at a job center, they don’t stick around for the day.

      4. I agree as well. Less parking — or at least more dispersed parking — seems like one of the more plausible predictions. We should see this at the airport first. Why spend all that money to park your car for a couple days? Just send it home after you are done. That is assuming personal ownership of cars. While the trend towards a “sharing” economy is a strong one, my guess is that there will still be plenty of people who want it both ways. They want to personalize their car — have it at the ready, with the car seat attached, the seat adjusted just so — while enjoying all the benefits of the new technology. This might mean more cars parked a ways off, away from the expensive lots. But unless we really get to the point where fewer and fewer people own a car, we will likely still have plenty of cars parked somewhere, just not where they are parked right now. For a trip to the ballgame, for example, it might mean a car parked at a nearby neighborhood, as sending a car all the way back and forth will still cost more than free parking.

      5. “They want to personalize their car … with the car seat attached”

        This is an issue that goes beyond convenience, as Martin has said. Either the driverless taxis will have to have car seats in some of them (sometimes more than one), or people will have to carry around car seats which they won’t do (they’re bigger and heavier than a bicycle helmet). We non-SOV advocates haven’t really come up with an adequate solution to this yet. It will become more of an issue as housing gets more expensive, especially 2+ bedrooms for a family, draining people’s pockets and pushing them to the suburbs.

      6. Besides car seats, there is also the issue of dogs. Little ones can easily be crated for car share; over 25 pounds or so a car share solution hasn’t popped up yet that I’ve seen.

  4. The real estate required by cars is enormous. Look at a map of downtown Seattle or any American city and look at the ratio of roadway to real estate. The ratio gets even more striking as you move into the suburbs and start to see surface lots and the buffers around the freeways. Even at my own house a quarter of my yard is driveway and a quarter of my house is garage.

    I was able to spend some time in Vietnam recently and experience their scooter culture. It was really interesting to feel how different commuting on a scooter feels from being in a car. You communicate with other drivers much more directly because when you pull up to a light you are literally close enough to converse with them. Horns are ubiquitous, but also hand gestures can be used to negotiate right of way.

    Scooters are easily stolen so they are usually brought inside the house or within a gated patio or yard at night. Houses occupy the entire street and but up against one another since there is no need of driveways. Streets also get very narrow, often accommodating a car going in only one direction since scooters can travel both ways in that width with no problem. And, especially in the city, apartments will be located down long branching allies which are basically just wide enough for one scooter to navigate.

    Self driving cars are an interesting development, but there are a lot of different ways that technology can go. What if we start to favor relatively cheap personal technology rather than ever more expensive and heavy automobiles.

    1. I found it interesting also that during the daytime the scooter parking lot became the sidewalk. In addition, their highway speed limit was 45. Point is, is that society also has to adjust.

      I liken the ‘war on cars’ as process change, where you have the death of something along with the birth of something. Self driving cars strike me as the bargaining phase. “But we can solve our problems and still have cars if we have self-driving cars”.

    2. Streets aren’t just for cars, though. We’d still need a lot of road surface for transit, biking, and walking unless people stopped traveling as much altogether.

      Let’s take 4th Avenue – call it 50 feet across.


      4 traffic lanes @ 11′
      1 narrow bike lane @ 5′

      Car-free future:

      2 NB bus lanes @ 12′ (need 2 lanes to allow buses to skip-stop for efficiency)
      2 cycle tracks at 8′ (NB and SB)
      2 median strips @ 4′ (1 between bus and bike lanes, the other between NB and SB cycle tracks)

      That’s basically the same amount of street.

  5. Driverless technologies for transit have been in use for years in a controlled, exclusive track environment. The most recent improvements are enabling automatic collision avoidance, and guidance on a path without a track.

    While higher speed travel is still not here for transit vehicles in mixed environments, rubber-tired driverless vehicles following defined paths at slower speeds are operating today. See

    The question is whether this is an end stage concept for 10 or 20 years or if Metro buses won’t need drivers by 2030. I could easily see a day when we will see lots of smaller, more frequent, driverless buses instead of the less frequent, articulated ones.

    Another application may also be automated vehicles that one could send home from a nearby station, or driverless shared-ride vehicles. Drop-off is growing rapidly. We may not care about park-and-ride by 2030 and we probably will need to redesign stations for more drop-off and pick-up in the near future.

    1. I agree. A lot of people have said the same thing. From a transit perspective, this could be huge. I asked the question, and I don’t know if anyone knows, but how much of a transit agencies budget is spent buying and maintaining a vehicle? What is the relative cost of the vehicles (are smaller buses a lot cheaper)? What about gas mileage — is it a major expense?

      If maintenance and gas mileage aren’t a big expense, I could see a lot more buses on the road in the middle of the day. Why have them sit around, if you can add service for a relatively small amount of money (the cost of gas and maintenance). If that isn’t the case, then I could see a lot more small buses in action. They might serve the remote places during rush hour, then cover the common routes in the middle of the day. That way the big buses are only used where and when it makes sense to run them.

      I could easily see a lot more smaller buses on the road, which would really help improve transit immensely. Add frequency, and the world is a very different place. Anecdotally, I know people who take Link, for example, even though it isn’t the fastest way home. It runs often, a connecting bus runs often, and even with the transfer, it is faster than waiting for the more direct bus. Same with RapidRide. It hasn’t become any more rapid, but frequency matters a great deal, and can lead to very high ridership.

      Your point about park and rides is a very good one. Right now, the hourly rental car companies as well as the new cab companies (Uber, etc.) don’t do a great job in the far off park and rides. This leads to a lot more driving, even to downtown. For example, I have a friend who works downtown and rides the bus or train every day. But he drives once a week to visit his dad in Lacey. He could take the bus, but the transfer would be very time consuming, or expensive. But an automated car at the other end would be great. He could take the bus from downtown to Olympia (and probably get their faster) and then quickly ride the car. The car would take him back, and he could reverse his journey fairly easily. Worse case scenario he rides the car all the way back home, which would be relatively expensive, but only because he has to pay the middle man. Their would be no more fundamental expense (there would actually be fewer miles put on a car).

      1. “But he drives once a week to visit his dad in Lacey.”

        I have run into this attending events at Arlington, Quilceda Casino, Mt Vernon, etc. I’d like to take a bus to the outermost station (e.g., Everett) and get a Zipcar from there. But Zipcars are few or nonexistent that far out, and even if they were at the P&Rs if more than two people wanted to use them simultaneously it would overwhelm the cars. So instead we have to drive the extra thirty miles from central Seattle, which is sixty miles round trip, or an extra hour of car use ($10).

      2. Generally speaking, labor costs are about two-thirds of a transit agency’s operating budget. Of course, while about 75-85 percent of those are drivers, a driverless bus will take more staff to monitor and maintain the vehicles. Large management centers with camera access to each vehicle will be needed and will be manned, for example. Some basic cleaning and those kinds of things will also be needed. So, at about the same cost, driverless buses could easily give us twice as much service as regular buses. Then, there is the sea change impact on demand — and that could enable us to have driverless buses three times more frequent than regular buses.

        On the other hand, the shared-ride concept may end up reducing costs to the point where driverless buses are obsolete in many settings. It may be that the driverless bus market ends up like a streetcar or sideways elevator, but not a viable solution to serve low-density areas.

        We live in interesting times..

      3. From what I’ve heard from a Metro operator friend or two, the expensive part of operating coaches isn’t the operator, its the per mile/platform hour of driving the coach.

        One of Metro’s meta service patterns is the storage tripper. There are a whole fleet of coaches that do one run from the suburbs toward downtown in the morning and one run from downtown to the suburbs in the afternoon. The operators then park the buses down at Central base in Sodo and they get on a bus run that is only for operators back to their suburbish base. In the afternoon a different set of operators start their day at the suburbish base, get on a bus that is only for operators down to central, pickup a coach and drive it to the suburbs.

        The operators get paid for all of the time, from checkin to checkout. Its cheaper for metro to pay the operators for sitting commuting, than it is to drive the coach.

    1. That’s kind of the hallmark of SOVs. My car is very convenient for me, but the infrastructure required to support my use of it is not necessarily the best investment of our shared resources.

    2. Like most things, it will likely be a mix. There will be some problems they solve that would be great for the commons (e. g. small buses running all day long) but there will be plenty of problems they cause (congestion in urban areas similar to what you see next to the airport).

    3. Id imagine we’d see endless streams of autonomous vehicles bumper to bumper traveling through timed lights. Good luck crossing the street where there are crosswalks, there wont be a break in the traffic. Think of a long freight train going 50 mph on every arterial, like that. All day.

  6. Great article Zach. Reminds me a bit of a Slog post (and a better one at that). Great last sentence (got a chuckle out of me).

    The ramifications of wide spread adoption of automated automobiles are profound, and probably represents the biggest advancement in transportation for quite some time. Every aspect of our transportation system will change. If your average car comes with a free chauffeur, then the dynamic of a private car changes dramatically, as the Jarrett Walker article stated. Cabs become much cheaper (got to hand it to Uber for their classic American business model — first screw over the employees, then replace them with machines). In many areas, congestion would likely get worse as taxi-cabs, and cars acting as taxi-cabs (dropping off or picking up people) clog the streets. Ever wonder why there are laws limiting the number of cabs? This is why. As with taxis, I would guess that the laws will eventually catch up with the technology (e. g. congestion pricing).

    As is being discussed up above, I think the idea of platooned vehicles, or automated vehicles in general being more efficient drivers is exaggerated. New technology can come along very quickly, but old technology sticks around a long time. Turntables, audio and video tape, land lines and cars without airbags are not that hard to find. If 90% of the cars on the road are automated, that still means 10% may prevent the perfect merge, or the bumper to bumper travel that some think is right around the corner. Besides, as you state, even if you did have all of that, there are limits to what the technology can do. Every software engineer knows that sooner or later you run across physical limits (processing speed, memory, bandwidth) and you have to adjust. Try leaving a crowded ball park and the limits of perfect synchronization become obvious. Even a top class dance troop, marching in perfect unison, can’t empty a stadium very quickly. You simply can’t fit that many people on the aisles. The same is true with the roads (as the gif clearly shows). There will be a lot of very important changes that come about as automated driving becomes more common place, but congestion relief is not likely to be one of them.

    From a transit perspective, the change could be enormous, but again, it has little to do with congestion. Anyone have any idea what various types of buses cost? I don’t. But assume for a second that buying and maintaining a couple 20 seat buses is not much more than buying a 40 seat bus. Once you stop worrying about the driver, everything changes. Every bus run has five minute frequency. Congestion is still a problem, but to many transit riders (myself included) the world changes dramatically. Why drive, if the only difference in speed is either the route the bus(es) take, or the number of stops? I would certainly ride transit a lot more often if I didn’t have to check a schedule, and knew that every transfer involved a tiny bit of waiting. Having a mix of buses (some small, some large) on the streets, day and night, changes the dynamic considerably. Suddenly our little town becomes very much like a big city. I might call a cab, or drive, but those are luxuries, and don’t save that much time (unlike today, when the difference is huge).

    Then there is the prospect of more dynamic transit, in less densely populated areas. Buying and running automobiles is not that cheap, so it makes sense to maximize their use. If the minibus I mentioned only has a couple riders on each trip, despite running every five minutes, perhaps there is a better way. I would expect to see more on-demand transit systems, like the one that Kansas City has just started up ( That helps solve the “coverage problem”, so that fixed route buses can focus on efficiently meeting demand.

    What I find bizarre is the notion that automated cars will somehow make transit obsolete. They will take away one of the big benefits (the ability to read or otherwise take your eyes of the road while traveling) but that is about it. Some dynamics stay the same — you can only fit so many cars on the street, cars, buses and trains are still expensive. But if most vehicles come with a free driver, then this would tend to favor shared, if not public, transportation.

    1. Platooning taken to the extreme could involve channeling autonomous SOVs into fixed channels once they get to sufficient density. For example, the vehicles could drive freely out in the suburbs and then be channeled into a tunnel system within the city. Vehicles like this could serve the same purpose as light rail for station to station travel, but when they get to the portal at the edge of downtown they don’t continue on fixed track, but spread out into the regular roads from that point on.

      1. “channeled into a tunnel system within the city”

        Can we build the second Link tunnel first?

        Also, this could be a potential future for the Deeply Boring Tunnel.

  7. I dunno. We already have cars with enough computer power that they are hackable by car thieves when standing still. When (if) we have self driving cars, platooned to the gills on I-5, all its gonna take is a couple of self driving cars without tip top anti-virus protection to shut Seattle down.

  8. Regardless of future technology we need road diets, dedicated transit lanes, wider sidewalks, protected bike lanes, transit oriented housing and businesses, etc. Everything to make the city safer and more enjoyable for households of all incomes.

  9. The push for driverless cars comes from SOV drivers. In the 1950s Santa Clara County voted against BART to put the federal money into expressways instead. That was the belief at the time, that a few highways would give plenty of capacity for everybody to travel by SOV long-term. Now they have clogged highways and people can’t get around. They’re finally building that BART thing they should have done fifty years ago. They also built a light-rail system that was intended to replicate the streetcars they had ripped out just 10-20 years earlier. But they didn’t understand the concept of where non-drivers travel throughout the day so the light rail doesn’t do much except take people from San Jose houses to Santa Clara office parks, and it doesn’t even do that very much. (Rush hour on VTA looks like Link at 8:30pm, and U-Link may have pushed the boundary further.)

    Everywhere in the country, and most Pugetopolans, don’t consider any mode except driving, and most of them don’t have any transit options that are reasonable given their home, workplace, and preferred shopping center. (Because the transit hasn’t been built and the land use is not amenable to it.) Or they might use transit in limited cases such as going downtown or to a ballgame, but not beyond that. Many of these people don’t want a non-SOV solution, they want driverless cars and electric/hydrogen/solar/hybrid cars to make driving viable long-term. The rise of driverless taxis complicates this somewhat, but the explanation is that some people want to own a car and keep their stuff in it, while others just want a driving service that avoids sharing the vehicle with strangers or being limited by stations and schedules.

    We, as armchair city planners and aware of where non-drivers want to go and when, look at other factors in how well the city functions and meets people’s essential needs and wishes to be at different places at different times. But SOV drivers don’t think that way; they think, “How can I go in my car to all the places I want to go to?” Or sometimes, “How can I Uber everywhere?” And, “That traffic is getting in my way; we must build highways to reduce congestion, and maybe driverless cars would help too.” The first premise is that only SOV modes and adequate parking are acceptable, and everything else must go around that. “Adequate parking” is 2.3 spaces per car or such: one at home, one at work, and a shared one at the supermarket.

    Jarrett Walker points out that even a civilization on a different planet would find that creatures take up a certain amount of space, they want to be different places at different times, an enclosed vehicle is larger than the creature, and the vehicle must be stored somewhere between trips. So if the aliens choose an individual-vehicle mobility system, it will have the same congestion/parking/space problems we do. If they choose a skateboard-like system, much less space is required. If they choose an individual flying-car system, that may give more space but The Jetsons shows that mass commuting would still lead to air highways (permitted travel corridors) and congestion.

    Jarrett’s focus on geometry needs to be shouted from the rooftops because SOV drivers don’t understand the concept, and the officials they elect don’t understand it either. They have been fed a Futurama/Brady Bunch vision of everybody driving easily on uncongested roads to everywhere they want long-term, and they think it’s the epitome of civilization, or at least the American Dream and birthright, and inconvenient issues like the space parking takes up doesn’t exist. The believe Le Corubsier’s vision of towers in the park with highways connecting them (even if they don’t know who Le Corbusier is), and the Broadacre City complement (e.g., exurban farm estates and small offices on highway exits), and a grotesque distortion of the Garden City movement (small satellite towns with a quarter acre for everybody; the original was more like New Urbanist row houses and small-lot bungalows laid out in streetcar suburbs).

    Driverless transit and non-fossil-fuel transit can of course leverage the same technology as driveless cars. And it’s much easier for a small fleet to convert, fuel them centrally, and they only need a subset of roads upgraded. Those factors suggest transit will change more quickly and dramatically than the nation’s car fleet. But it’s also limited by budgets, adversion to taxes, legislators that neglect it, and local politicians that don’t grasp it (e.g., ST not prioritizing driverless trains using existing technology, and a baseline premise that 100% grade separation is necessary anyway both for that and other reasons.)

    1. I agree “The push for driverless cars comes from SOV drivers”. Maybe if they rode a bus, they’d get most of the benefits of driverless cars without encouraging highway expansion.

    2. “They have been fed a Futurama/Brady Bunch vision of everybody driving easily on uncongested roads to everywhere they want long-term..”

      Looks like average the car commercial on TV tonight. We’ve all been fed that.

      Wouldn’t it be fascinating if car commercials actually had real traffic?

  10. Maybe I’m following my little soldier into battle, but I doubt seriously multiple transit geeks are all that happy about the idea of driverless cars clogging up roads, putting pressure to pave over more farmland & wooded land & front lawns. Good grief.

    GEOMETRY, PEOPLE. Figure out how big you are sitting down are versus say a Ford Mustang in surface area. Good grief. Surface area! Not that hard to figure out!


    1. See if you can raise awareness on this:

      My New Article on Transit’s Space Efficiency (2016; one Washington Post article and one southern California government report)

      the photo that explains almost everything (updated!) (2012; car/transit efficiency in pictures)

      how universal is transit’s geometry? (2011; alien civilizations)

      The most urgent issue is this: “”Personal Rapid Transit,” or small demand-responsive buses, or driverless cars that work like taxis, will never, ever, ever substitute for surface transit in high-demand urban settings,” (second article). Some people are advocating we don’t build any rail or BRT or even replace bus fleets because widespread driverless cars are immanent in the next 3-5 years and will make transit obsolete. This is dangerous if it leads to not doing multiyear projects, because when the time comes and people realize it’s false they’ll still need several years to build them, and by then the population will have risen and bus overcrowding will be more acute.

      I also worry about chasing after experimental train technology. Light rail has been proven for a hundred years, is off-the-shelf, and its remaining problems are well known. Some driverless technologies like Skytrain have been proven for thirty years. But production fleet decisions must not be based on technologies that are still in testing or the lab, because that’s depending on something that may not perform as predicted or be immediately obsolete.

      1. This is my biggest criticism of the Hyperloop. We can buy an off the shelf train from any number of manufacturers which runs fully twice as fast as anything we have going in this country. The costs and risks of construction are well known and we can probably come up with good ridership projections and so on. Or, we can not do that and bank on some entirely new technology disrupting the entire technology sector and building that instead. Assuming we need a high speed rail line between SF and LA I would strongly opt for going with proven technology. Yes, we risk wishing we’d built a Hyperloop, but more likely we’ll be riding the train we knew we could build for decades before they figure out how to get the Hyperloop capsules to turn efficiently.

        So, yes, by all means, let Musk pour billions into developing the Hyperloop, but invest our transit dollars in proven systems we can buy now.

      2. There are a number of reasons the “hyperloop” never got off the ground when they first started playing with the idea in the 1830s, and in successive proposals for various similar concepts.

        One of the significant issues is fresh air intake for the passengers. It’s a sealed can surrounded by vacuum, so no traditional off the shelf solution exists for fresh air intake. Current standards in passenger cars are something along the lines of 6 cfm per passenger, and closer to 10 or 12 for standard building systems.

        It’s going to need a space capsule fresh air system, only able to deal with a crowd of what? 100 passengers?

        Then there is the issue of keeping the tunnel sealed while getting passengers in and off the train. Vacuum is extremely difficult to keep sealed from atmosphere.

        It isn’t that it can’t be done, but the fact is, there are reasons why things stay research projects for 150 years.

        Driverless cars are a different matter. We have the computing horsepower to do it. We don’t have the ability to teach computers about abstract concepts though. “That chair isn’t tied in to that pickup very well, so I should stay back further” is an abstract concept of something that might happen, not something that has happened and needs a response.

        So, there is no question we can do driverless cars, but I do question their ability to make any sort of vast revolution in road utilization or safety or the various other claims.

      3. I, like many, don’t like the idea of being shot in a tight tube at mach speed. I’m amazed someone like Musk would throw good money after bad.

      4. Musk put $0.00 into the Hyperloop. Musk isn’t stupid. He let other people waste their money on it.

  11. The robots will save us from ourselves. All hail our future robot overlords.

  12. Deciding now which technologies are best for the future is an act of hubris.

    Allow people to work on driverless cars, SkyTrans, Hyperloops, hybrid buses, crotch rockets, transporters or whatever suits their fancy. Whatever fits needs best will probably flourish.

    Who would have predicted that cars would be overwhelmingly more popular than interurban trolleys? Or that aircraft would make ocean liners obsolete for transportation?

    And so quickly!

    1. +1 Al’s comment on external costs – but there’s more that lies outside of market solutions. Effective transportation requires extremely large-scale infrastructure commitments and land-use decisions, and alternatives often can coexist only with a LOT of friction (eg, problems of grade separation). In the US we went in the direction of private automobiles and roads to support them, much more decisively than most other developed countries. And that has become a self-fulfilling prophecy. Until there is a critical mass of transit infrastructure built, a large fraction of individuals (including Zach!) are more or less forced to drive cars. In economic terms, this is a problem of network effects, or increasing returns

      Even with pricing of (individual, marginal) external costs, it still requires a very deliberate, significant, and costly change of direction to make enough of a critical mass for the transit alternative to be competitive to the point that “whatever fits needs best will…flourish.”

      Of course, large transit investments can be made well or poorly, and automated driving will probably have a sizable impact on what constitutes good policy. The convenience and other advantages that make cars attractive even in transit-rich countries will tip the balance against efficiency more often if automated driving makes cars + roads more efficient. (Though presumably automated driving will make rail and buses more efficient too!)

  13. It may eliminate parking in the densest and most high value land places but I suspect there will be giant layover parking lots in low value areas (probably poor communities and industrial areas) to hold excess cars during non-peak times. Not all that dissimilar from bus yards/bases. I think low land value places just outside high land value downtowns… oh like SoDo and Interbay.

    If this isn’t the case, there will be a lot of long distance empty vehicle deadheading that doubles traffic.

  14. I’m curious how this works for pedestrians. Sounds like, at least initially under current laws, the cars are pretty much guaranteed to stop for pedestrians when they enter the street in a crosswalk (or really anywhere) as its an “obstruction” in the street. Riders in these autonomous vehicles will not like that one bit. I suspect allowing the pedestrian the right of way in a crosswalk wont last long under the new autonomous system. We very likely will be refighting the same battles for the streets that were fought 100 years ago between the newly emerging cars vs everyone else. We are kind of fighting them now with livable streets but it will be nothing compared to what is coming… it will be much more fundamental.

    1. All the same, I’m definitely looking forward to those few years when I am able to cross the street the way state law says I should.

  15. Zach – Nice article! Showed it to other out of town transportation folks (who don’t read STB) and they liked it too.

Comments are closed.