Autonomous vehicles and urban mobility

A discussion of the potential impacts of autonomous vehicles on urban transport.

There has been much written about Google, Inc’s recently announced forays into autonomous vehicles, including the recent decision by the Legislature there to legalize operation of the vehicles. While Google’s efforts are still in the experimental stage, they appear to represent a significant advance in the state of the art, and have inspired much discussion on What It All Means for urban mobility in the future.

Some transit critics see the autonomous vehicle, whether privately owned or operated as driverless taxis, as displacing public transit–Randall O’Toole was quick to predict that the technology would make rail transit obsolete. Others see it as providing transit an opportunity to expand its operations. While it’s too soon to make any firm predictions, it’s an interesting subject to discuss.
Railroads and automatic operation

In ground-based transportation, use of automated or centralized systems to assist with vehicle operation has long been most prevalent on the railroads. Trains, especially long freight trains, have long had a fundamental safety issue: the safe stopping distance frequently exceeds the driver’s line of sight. This is true both because of the length and weight of many consists, and due to the low coefficient of friction between rails and wheel. Unlike road-based vehicles, which can more or less proceed safely based on local observation, trains cannot safely operate at speed in this manner.

The earliest way of preventing train collisions was a simple one: the timetable. In addition to the other benefits of scheduled service, railroad schedules ensured that no more than one train would be in the same section of track at a time. Other crude signalling mechanisms, such as explosive charges deposited on the rails, were employed to deal with breakdowns and such. Track reservation by timetable was highly unreliable–accidents were common–and inefficient.

As technology improved, mechanical and later electric wayside signals were developed to permit finer-grained mutual exclusion of train segments. As communication technologies improved, central monitoring of signalling (and centralized operation of switches) became commonplace. Many freight systems, and much of MAX, use a modern version of this approach known as Automatic Block Signalling. In addition, MAX trains are equipped with safety devices which will stop trains that attempt to proceed past a red signal, for example. (MAX FAQs is a great resource for those wondering about MAX signalling).

ABS has its limitations, however, and some railroads go further and deploy more advanced command and control technologies, which go by such names as Automatic Train Control (which replaces wayside signals with cab signalling, and the fixed blocks of ABS with continuously-moving blocks around trains) and Positive Train Control (think of it as automated air traffic control for railroads).

None of the above technologies replace the train engineer, however. A few transit systems have gone further and implemented various types of Automatic Train Operation, which takes over some or all of the duties of the driver. In systems such as the Washington Metro (which featured automatic operation until the 2009 accident that killed 9 people), ATO was used as a capacity-enhancement measure and only during rush hour; drivers are present on the trains at all times. Vancouver’s SkyTrain, on the other hand is completely driverless; under normal circumstances there are no crew members aboard. SkyTrain vehicles even lack cabs; passengers may sit in the trains’ nose should the choose.

Both the Metro and SkyTrain’s ATO systems are limited in many fundamental ways, however. They lack the capability to detect obstacles (or potential obstacles) on the track, and thus require segregated rights-of-way. (Both use third-rail power, as well; making the tracks unsafe for trespassers even if no train is around). And neither is capable of autonomous operation–both require constant communication with the ATC system.

Autonomous operations

An important detail of the technology being developed by Google is that it is autonomous; Google’s driverless cars do not require any special infrastructure outside the car, other than ubiquitous services such as GPS. Cameras and other sensors, backed by complex image processing, detect the location of the pavement, the presence of obstacles or other vehicles, traffic control devices, and such–the onboard control system functions much as a human driver would. More reliable interfaces are possible, such as car-to-car and car-to-signal communications, but these aren’t required for operation–the autonomous cars can navigate ordinary roads.

Some have suggested that assuming the technology becomes widely available, that the importance of public transit (specifically, large vehicles like busses or trains designed to carry dozens or hundreds of passengers) will be diminished. To test that claim, it is useful to examine the disdvantages of cars (for passenger travel), and see how driverless operation will change things. Disadvantages of driving (both for motorists and for society) include:

  • Driving itself is inconvenient. Some motorists are auto enthusiasts–they love driving, and working on, automobiles, and often purchase cars based on their “style” or on their handling and performance characteristics, rather than on more functional criteria. For most motorists, though, driving is a pain in the butt–a chore which is an often-ignored cost of car ownership. Who enjoys sitting in rush hour traffic with their foot riding the break? Who enjoys a twelve-hour drive to Sacramento? Who enjoys circling a block looking for parking? Also, there are some people who cannot (legally) drive, due to physical disability, lack of license, or revocation of the privilege by the law.
  • Cars are dangerous. Automobile collisions kill tens of thousands of people every year in the United States, and injure thousands more. In most cases, driver error is to blame. In addition, automobile traffic, particularly high-volume or high-speed traffic, is detrimental to the pedestrian environment; highways, in particular, act like walls which pedestrians are unwise to cross.
  • Cars are expensive. Owning or operating an automobile costs a lot of money–they are expensive to purchase, expensive to fuel, expensive to maintain, and expensive to insure. In some locales, they are also expensive to license, operate, and park. Likewise, arrangements by which one can use an automobile without owning are either expensive (car rental, taxis) or inconvenient (informal carpools).
  • Cars consume a lot of space. A single-occupant vehicle (SOV) requires, on average, about 100 square feet of real estate (including some margin outside the vehicle’s footprint) to carry around the person it contains. That’s a 100-square-foot patch which must move down the streets at speed, while avoiding encroaching on any other 100-square-foot patch, and which must then be parked when not in use. This consumption of space is a major problem in dense urban areas; and a common solution–less density–makes things inconvenient by spreading everything out further, requiring cars (or other mechanized transport) for trips that might otherwise be easily accomplished on foot.
  • Cars consume energy and produce pollution. Along with the area consumed by cars, they weigh a lot. A 150-pound commuter lugging around a 3000-pound automobile simply to get to work is a profound waste of energy. And given that most cars today are powered by internal combustion engines, this results in a lot of filth out of the tailpipe.

Well-used public transit addresses each of these issues. Transit passengers need not operate the vehicle; transit is operated by professional drivers, and some transit already is automatic (accidents still occur, but it is rare for transit passengers to be killed about a transit vehicle); transit rides are available at low cost (and a transit pass is far less than the annual cost of owning even a beater car); and well-used transit requires a lot less space and weight per passenger. (That said, walking or biking does even better than transit on these last two issues).

What about driverless cars?

  • Driverless cars solve the “driving” problem completely–rather than driving to work, one simply rides to work, and can use the time for more productive pursuits than operating a motor vehicle.
  • Assuming that the technology reaches its potential, driverless cars ought to be much safer than human-driven ones–for passengers, pedestrians, and other motorists. Computers have much faster reaction times than do humans. Computers don’t get drunk or tired or have bad days. Computer control systems won’t engage in antisocial or aggressive behavior (assuming proper programming), and don’t suffer from the need to demonstrate their machismo behind the wheel. And computers can easily be programmed not to disregard traffic controls, in particular speed limits. While automatic vehicles won’t eliminate accidents, automated cars should reduce this carnage. In addition, if and when driverless vehicles become commonplace, it may become practical to build lanes or roadways on which manual operation is restricted (commercial license holders only) or prohibited altogether–and maybe manual car operation itself might be greatly restricted, with only highly-trained professional drivers being permitted behind the wheel.
  • Driverless cars probably won’t reduce the expense of car ownership; and may increase it, depending on how much of a premium is charged for the control system. Insurance rates will likely go down for cars operated only in automatic mode, but other expenses are not likely to be greatly affected. However, driverless cars may make it far easier for persons to have access to cars without owning one. Taxis, presently, are notoriously expensive (and often hard to find), due to the need to pay the driver. (And taxi drivers, despite the high fares, generally earn a pittance). Driverless taxis, on the other hand, will likely become ubiquitous, and arrangements such as “share taxis” might become more practical. Driverless automobiles will also make other arrangements such as car rental and car-sharing more practical. The delivery problem (how to get the car from the garage to the user) will go away, as might concerns about liability and insurance (again, assuming that manual operation of shared cars is not permitted) which presently frustrate car rental. (That said, Hertz will still try and sell you worthless collision damage waivers…)
  • Driverless cars will help greatly with the parking problem, even for unshared vehicles, simply by permitting parking to be relocated away from the home or jobsite. (Shared vehicles might not need to be parked at all, other than overnight). Rather than having distributed parking all over the place (with valuable street real estate consumed for on-street parking, and each business with a parking lot intended to handle their individual peak loads), parking could be centralized at an out-of-the-way location. When someone goes to work, they are dropped off at the office, then the car travels to a central garage for storage (having already reserved a parking space electronically); at the end of the day, the car is summoned, picks up the commuter, who then heads home. The effects of driverless cars on congestion are a bit less clear. If a roadway is reserved for driverless cars only, the carrying capacity of the road will increase significantly; and by solving the parking problem one also solves the circling-the-block-looking-for-parking problem. However, the need to travel to and from central garages might increase congestion, particularly if centralized parking is concentrated in a particular space.
  • Addition of control systems shouldn’t materially change the weight of vehicles; however fuel efficiency should improve for several reasons. Reduction of congestion and of aggressive driving habits (which often waste fuel) will improve fuel efficiency assuming no changes in the fleet; and anything that can be done to encourage carpooling will have a corresponding reduction in fuel consumption and pollution.

    A more important advantage might be greater fleet specialization. Right now, many commuters get to work in large vehicles such as SUVs which are particularly inefficient at transporting single occupants (with minimal, if any, baggage) to and from locations on the improved street network. However, these commuters may have hobbies or other circumstances which require ownership of a large vehicle–and at that point, it’s more cost-effective to drive the Expedition to work than it is to purchase an additional vehicle for the commute. By reducing the barriers to car-sharing, it may become more practical for more motorists to deploy appropriate vehicles for differing circumstances: a family could own a small, fuel-efficient (or electric) car for commuting, and then rent the gas-guzzling SUV for trips to the lake or other circumstances where the greater power, size, or clearance is needed.

How will this affect transit?

Many transit critics assume an auto-normalized environment–that the natural state of affairs is for low-density, auto-centric development, and that transit is only useful or effective for specific exceptions to the norm (high density, transport for the poor), often as a result of government interference. A common assumption is that people intensely dislike both high-density living, as well as travel in large communal vehicles–and that any way in which automatic cars can alleviate the problem with driving, will result in a mode shift away from transit back to the automobile.

For some people, this is certainly true–and I expect that certain applications of transit will become less prevalent. In particular, the express bus–which appeals to commuters who dislike driving in rush-hour traffic (though don’t mind being stuck on it while riding a bus), and/or wish to avoid parking downtown, but prefer to user personal cars for trips other than the commute–may disappear. Most transit agencies won’t consider this a great loss–express bus is an inefficient service to provide. (One concern might be a loss of political support for transit from suburbanites). Likewise, other arrangements may become practical for the frequently-empty social service routes through suburbia–again, a service which is inefficient to provide.

However, in dense urban areas, where stowing and parking a car is frequently expensive and difficult (regardless if manual or automatic), I don’t expect automatic cars to displace transit, simply due to the space aspect. A collection of small cars, even all running on autopilot (and packed onto the roadway like sardines), simply lacks the people-moving capacity of a rapid transit line.

More importantly, autonomous operation will provide advantages to the transit system as well.

Autonomous vehicles can reduce or eliminate the “last mile” problem associated with transit. Rather than needing to build park-and-rides to attract suburbanites (who live in neighborhoods ill-suited to quality bus service) to rapid transit, one can simply ride to the train station, get on the train, and have the car return home (where it can then be used by other members of the family); and repeat the operation in reverse for the commute home. This not only reduces the need to consume real estate to store cars during the day, it can reduce the number of cars a household needs. (And easier carsharing can eliminate the temptation to have a “spare” car on hand in case one breaks down).

But the most important factor to consider is this: Transit vehicles may become autonomous as well.

Right now, the biggest expense for transit systems in the developed world is labor. TriMet’s service levels are not presently limited by the amount of rolling stock or by the capacity of the infrastructure–the agency has plenty more busses and trains in its garages than it currently operates, especially at non-peak times. They’re limited by the agency’s ability to pay for drivers. If that factor is eliminated; we could easily have frequent service throughout the entire network, making transit a more realistic option for a larger number of people. (Obviously, this will require throwing hundreds of drivers out of work–a factor which should not be overlooked in these discussions).

Furthermore, and this should make the libertarians happy, autonomous transit may actually become a profitable enterprise rather than one which requires a public subsidy; with the result that private operators and not TriMet or C-TRAN run many (or even all) of the mass transit services in town.

Final thoughts

The production and deployment of autonomous vehicles, including mass transit vehicles, may bring about significant changes to how transport is arranged. It may increase the number of SOV trips (albeit making them more efficient) by eliminating some of the disadvantages of driving and vehicle ownership; but at the same time it might actually reduce the size of the vehicle fleet. It may result in a shift away from inefficient transit services, and better service in the urban environments where transit thrives. It may even result in a fundamental shift of who provides transport services to the public.

My suspicion, however, is that widespread use of autonomous vehicles will be a net win for sustainability–even if fundamentally alters, or even eliminates, the current ways that public transport is provided. When couples with the pressure of rising energy prices, autonomous vehicles may shatter some currently-held assumptions that presently abound in middle-class American society, such as the belief that households need one car per driver; and it may make urban living more attractive and available as the size of the fleet is reduced. It may end up greatly reducing public transit as we know it today–perhaps limiting it to operation of high-capacity corridors which require extensive infrastructure. But speaking for myself–my interest is not in defending the turf of TriMet, or of any other particular agency. My interest is in providing efficient and sustainable mobility to the people, and in particular in alternatives to the SOV.

But my crystal ball is probably no more clear on this topic than that of anyone else.


19 responses to “Autonomous vehicles and urban mobility”

  1. I think this technology should be implemented on some new fangled, ultra modern, high tech city bus system. This would probably double the ridership overnight, making other modes unneeded. [Moderator: Personally-directed remark deleted–ES]

  2. Thank you, Scotty, for your work in putting this altogether.

    Some initial thoughts:

    1. Some multiple vehicle households may see AVs as a way to reduce the total number of vehicles owned while still holding on to personal mobility. Even with $8 gas it might still be cheaper to deadhead AVs among household members than to pay depreciation, insurance, upkeep, etc. on additional vehicles. This would increase VMT.

    2. It could be cheaper for a downtown Portland worker to send the car deadheading back home after being dropped off than to pay high parking rates — currently around $150 to $200 per month. This, too, would increase VMT and add cars during peak commutes, albeit in the reverse direction.

    3. While it should be a lot cheaper for bus AVs to operate than manually operated ones, there probably still won’t be many suburban situations when they could compete effectively with strategically parked small rental AVs. If they’re plying regular routes, they’d still be running around mostly empty and eating up money; just not as much as the current fleet.

    4. Limited express buses could survive, particularly with super-HOV lanes or dedicated guideway, at least during commute hours. The key would involve the relative cost and time with small rental AVs.

    5. We can expect political pressure to eliminate or at least greatly reduce payroll taxes. I think the public might continue to support them to some degree if they were seen as being effective in decreasing congestion. Mostly empty transit vehicles would have to go and only the most heavily used services would survive.

    6. Public policy will determine whether AVs are boons or banes. Laws and ordinances would need to encourage, or at least not discourage, trip sharing and do the opposite for deadheading, especially during commute hours. Examples of “bad” law include the inability of suburban taxis to pick up fares at the airport and the effective monopolies of transit agencies in their districts. With good laws, Zip car/taxi AVs could go a long way towards allowing people to live without owning cars, or at least minimizing the number owned.

    7. Far more space is used by a vehicle for its parking than for its traveling. The space at home typically is used rarely by other vehicles. A parking spot at work is comparable, especially if there’s only one shift. A parking lot at a house of worship may only be filled once or twice a week. The space in front of Freddy’s is used by 30 or 40 vehicles a day at most. A 100 square foot block on an I-5 traffic lane is used by 20,000 different vehicles daily. AVs have huge potential for reducing the need for parking in high density areas. This would significantly reduce the cost of high density development while allowing for more “people friendly” streets.

    8. We can expect the heyday of costly fixed guideway projects to end within the next decade or two. It will be much more difficult for the Metro/TriMets of the nation to demonstrate a real “need” with acceptable cost benefit ratios. Heavily used existing guideway will continue to be used, but questionable lines – LO streetcar, WES, Red, etc. are likely to be abandoned.

  3. I’m surprised it’s starting to look like in the US we’ll be seeing this technology on actual private vehicles first, before things on rails. It seems the trains would present less difficulty, all you’ve got to do is watch for obstructions. Trains take a long time to stop, but a computer should be able to do better than a human, especially considering it can take data in from a multitude of cameras/sensors laying ahead.

    There are probably certain types of collisions a human might be better able to predict than a computer, but when I think of MAX injuries and deaths that happen, I think of quite a few that an automated system might be able to handle better.

  4. And, why not remote operation? I bet one driver could operate a number of trains safely, especially if you consider you can tack on automatic slam-the-brakes and whatnot systems quite easily once you’re driving by wire at a distance. More productive, and the job should be safer in a nice air-conditioned office in a comfy chair, with the ability to hand over control to somebody any elsetime one needs to take a break.

  5. Taxis, presently, are notoriously expensive (and often hard to find), due to the need to pay the driver. (And taxi drivers, despite the high fares, generally earn a pittance).

    A nit, but still. This is certainly legitimate in Portland. In the last few years I’ve been in Chicago, Boston, DC and several cities in the UK. Unlike Portland, in those cities cabs are cheap and ubiquitous. How else would four guys get back to their hotel at three in the morning after a night of Guinness and Black Bush?

    In Portland, taxis are expensive and require a phone call or a rare cab stand. This is not the standard anywhere else that I’ve seen.

  6. I would almost suggest that the cities with bad taxi service outweigh those with good service–the ones with good service will generally be large and dense (like the ones you mention) or full of tourists (Las Vegas comes to mind).

  7. Supposing for a moment that at-grade autonomous transit vehicles (perhaps just rail vehicles at first) can be developed to the point where they are safe and accepted by the public, one argument, as made above, is that costs could be saved by eliminating the operator.

    However, one could also view this as an opportunity for the reallocation of resources.

    Instead of what we have locally today, one operator per vehicle and no dedicated additional staff (fare inspectors and security move from vehicle to vehicle, the likelihood of having one on any particular trip is low — WES is the exception with a full-time conductor), we could reallocate personnel from operator positions to being stationed in the passenger areas of the vehicle, to assist with boarding, verify fares, discourage disruptive behavior, etc.

  8. I have trouble believing that even Google can design an autonomous car safe enough to drive on city streets. Sure, operating on grade-separated highways is easy enough, but once you have pedestrians, bikes, animals and bad drivers around, things get very complicated.

    It would make much more sense to develop fully automated rapid transit, then automated local buses, before trying automated personal vehicles.

    And this is the reason that I think automation would be great for transit. Once you can automate the bus, you would be crazy to pay for someone to collect fares. Finally, we could have POP (proof of payment) on every bus and train, as is done in all the good transit cities, with subsequently faster and more reliable trips. With automation, BRT really would be cheap, and streetcars could run every 5 minutes, 20 hours a day, for the same price as our current crummy service.

    If the technology could expand to automated taxis, transit companies would save a fortune on paratransit costs (which might otherwise kill the system in 10 to 20 years based on current trends), and people could easily get to destination that are off of the high-capacity lines.

    This makes “commuter rail” or regional trains, with stops every 2 to 5 miles, much more effective. You can bike or take an automatic taxi to the nearest station, then take a fast (110 mph) train until near your destination, and then get another taxi (or a rental bike if it’s close) to complete the trip. And the trains could be run very frequently, without the need to pay for operators, making up for the usual disadvantage (waiting time) that really kills commuter / regional rail systems in the suburbs.

    Sure, automation would make private vehicles nicer, as well. But the costs of driving and parking are a real problem for many people, and for every dense city, and automated electric cars won’t solve most of the real problems. Automated transit, on the other hand, has the opportunity to revolutionize transit in moderate-density areas, like Portland.

  9. Cynical prediction: automation will be awesome right up until the first time software error kills someone (no matter if it’s less common than human error), and then we won’t have it anymore.

  10. Joseph E-

    It does seem improbable that we could have AVs on city streets — except that much of what would be required has already been done. Think of Google’s over 100,000 miles of testing, Volvo’s S60 auto braking if it detects a possible collision with a pedestrian at speeds up to 22 mph, etc. The main thing is that nobody involved with the development of these things is publicly identifying any specific technological hurdle that looks insurmountable. The primary barriers seem to be social and legal.

    As Scotty has identified, there are a lot of things involved with rail transit control systems that are not present with personal vehicles. So it’s a real question as to how much to invest in rail automated systems for what amounts to very few potential customers or to go for the hundreds of millions of potential sales in this country alone for personal vehicle systems. I understand that the made-to-order system for Vancouver takes a lot of specialized maintenance. How many transit systems would really want to get into that?

    I think that there is more potential for transit buses which would have systems very close to those of personal vehicles. The big problem for transit is that it still operates on a centuries-old model of fixed route, fixed stops, fixed schedule, and fixed prices which will not be able to compete with AV taxis or car rental systems except in very few specific situations:

    1. Heavy congestion where the public would be willing to subsidize transit to keep that congestion at a minimum — think commute hour and circulators.

    2. Specific grade separated routes where transit would be competitive with personal vehicles outside of commute hours — think of OHSU’s aerial tram, a possible high speed transit tunnel for the SW project, a Lake Oswego-Milwaukie MAX extension, etc.

    It will be the AV taxis which have the best shot of giving people the possibility of personal vehicle mobility without having to be able to afford their own cars or requiring all that land devoted to parking space. In fact, it would seem likely that expensive parking space would be the first thing to go once AVs become commonplace.

    Aaron G-

    There is an awful lot of reality in what you predict. We can expect car manufacturers to go through the nth degree of testing before putting these things on the streets. The hope would be that they can amass a stellar safety record in comparison with driven vehicles even including that accident.

    There is also the possibility of malevolent hacking causing problems. The systems will have to be extremely resistant to such attacks. For this reason, I don’t see inter-vehicle communication (such as Ford is working on) as being a likely route to success. It’s bad enough to rely on GPS.

    I’m also concerned with half-way measures such as what VW is about to introduce with their TAP (Temporary Auto Pilot) system. We really don’t need gadgets that will lull drivers into the false impression that they don’t have to pay attention to what’s going on.

  11. Cynical prediction: automation will be awesome right up until the first time software error kills someone (no matter if it’s less common than human error), and then we won’t have it anymore.

    This brings up a psychological issue I’ve wondered about… In the event of accidental death where another human is responsible, that human can say “I’m sorry” (and sometimes mean it), and/or that human can be made to suffer for their negligence (public shaming, legal restitution).

    Both of those things may have a psychological impact for the friends and family of the deceased.

    When an autonomous computer makes an error which leads to death (or, is involved in a situation which unavoidably lead to death no matter what reflexive actions could have been taken), that computer cannot say “I’m sorry” and mean it, and no amount of punishment of the computer can make anyone feel better. (The laser printer scene from Office Space notwithstanding.)

    Sure, next-of-kin can sue the insurance company governing the autonomous vehicle (and it will be safe to assume that such vehicles will always do an online check before operating that the insurance policy is in effect). And that will settle financial matters according to the law. But if nobody feels “sorry” about it, can people move on as easily?

    Perhaps the opposite will be true: The removal of someone to blame (after the inevitable tide of lawsuits targeting vehicle designers ebbs) will allow for the “well, what can you do?” response and people will go on with their lives.

  12. With AVs, I imagine the automakers are going to have a bit more concern as to whether or not the car is being correctly maintained. Also, what happens when inevitably it’s fifteen years later and you’ve got this first generation of AVs that simply aren’t as error-proof as the newer ones, or don’t talk some fancy new car-to-car or car-to-cloud protocols that increase efficiency/safety?

    Are there any indications we might be moving away from true personal ownership of vehicles towards a more managed situation where you’re paying to use/lease/rent the cars?

  13. This is an excellent summary of the some of the impacts of driverless cars.

    If anyone here is interested in keeping up on the latest developments with driverless cars you might be want to visit

    It’s a site I’ve put together to explore these issues in detail. Please feel free to stop by and add your thoughts to any of the posts.

  14. Autonomous vechicles could be parked in a much denser fashion than human-parked vehicles. look at valet parking lots, where they pack in 2x the cars or more. with all vehicles automated, someone just calls their car and the cars around move out of the way for it. This is assuming that private ownership is the way of the world in the future

Leave a Reply

Your email address will not be published. Required fields are marked *