Brad Templeton Home

Main Page

Brad Ideas
(My Blog)

Robocar Blog

My Forbes Articles

(Book from 2008)

The Case



Car Design






Future Transit













Urban Planning



End of Transit




Sidebars: Charging

Speed Limits

NHTSA Levels

Valley of Danger


Google Cars

Personal Rapid Transit (PRT) vs. Robocars

Personal Rapid Transit (PRT) vs. Robocars

Many may find similarities between my talk of personal vehicles and the end of transit to the long-time transit proposal known as PRT, or Personal Rapid Transit. PRT is a system where self-driven cars ride around on a (hopefully inexpensive) track, going point to point without intermediate stops. Typically they involve elevated tracks or monorails, and certainly sound better than mass transit. They tend to be small, light vehicles that are more energy efficient than cars or most other transit.

Indeed, I thought a fair bit about PRT before judging that robocars were the future. I wrote plans and support for hybrid PRTs that could be human driven once they left the guideway, what is known as "dual mode" among PRT designers. The PRT vision is in fact a simpler version of the robocar vision which uses the elevated tracks or rails as a substitute for software that can navigate cars on ordinary streets. This makes sense before we reach the cusp of software which can handle that -- but even without this objection, the amount of PRT in the world today is tiny, if there is any true PRT out there at all.

One can also argue that robotaxis are a realization of the PRT vision, though not as most PRT advocates saw it. Most PRT fans don't tend to agree.

I see people say that "PRT is something we could build now, while Robocars are perhaps 10-20 years away." Yet PRT was also "ready to deploy now" in the 70s and for reasons that are much argued about, never really happened.

PRT requires lots of construction and expense, generally uses cars that are all the same size (2-3 people) and still only goes to where there are stations. At this point, making PRT happen clearly requires lots of political will, a will possibly also applied to robocars. You must fight every NIMBY (or NIMFY) who opposes an elevated track going by their bedroom window or other ways the tracks may var the view.

There are two proto-PRT systems in place, both of which use cars on tires, but have either dedicated tracks or magnets embedded in the pavement. ULTra PRT at Heathrow uses an elevated track. With slow speeds and only 3 stops, this is not anybody's idea of a full-blown PRT, but it is a beginning.

Robocars have almost all of PRT's advantages with a few exceptions: Travel on steel rails is more efficient (though noisier) than rubber tires, but this is, I think, dwarfed by the fact that roads are already everywhere while tracks and rails would still need to be built. New dedicated right-of-way is of course wonderful for any transportation system, and would give PRT a way around congestion, but the money to build such right-of-way could also do wonders for robocars in so many ways. Indeed, at a minimum it could build private ROW for robocars.

Rails can be built more cheaply than roads, but this only is of value if we can get away without roads completely, and that's a long way away, if it can ever happen. Users demand a set of solutions to "the transit dilemma" -- providing a good, fast trip for every two points I want to travel between. In particular, the delivery of large cargo to any location still demands a road network like we currently have -- it is beyond the ability of small PRT pods, which in typical designs hold 2-4 people, and beyond the abilities of PRT guideways deliberately designed to handle only light loads.

One special advantage rail-based PRT could have is the sending of electric current via overhead wires and/or rails. This allows for very efficient vehicles with no batteries, and that is a major win. However, I am not convinced that, should it be so compelling, that robocars could not be designed to make use of wires and rails to get power in areas where such infrastructure is involved. They would still run on other power when away from such rails.

The robocar's ability to leave the system and still be automated is the overwhelming advantage that may doom PRT. People demand vehicles that can take them anywhere, from dirt forest roads, to driveways and parking lots. They could be happy with the ability to transfer to more efficient vehicles on major routes, or even to have hybrid vehicles that can both drive a special rail track and also go on ordinary roads, but Americans have never been satisfied with systems that just go along designated routes to desginated stations -- not even in New York, the most transit-using city in the country.

To answer this, many PRT plans call for "dual mode" vehicles which can leave the system under control of a human driver. These are either rubber-tire PRT systems, or pods with both rail wheels and tires, or multi-part vehicles. Needing human control they lose many of the advantages of robocars -- self-parking, self-refueling and the empty-vehicle moves needed for sharing and taxi service.

Robocars have so many other advantages in marketability over PRT that I really see little contest. The only thing that makes it a contest is that robocars do not yet exist and need major research, while PRT barely exists but could be built if there was the will.

  • No new infrastructure is needed. That's incredibly huge.
  • Robocars will take you from anywhere to anywhere that pavement or even dirt trails exist. No limitations based on where stations or lines are.
  • Robocars could come in all shapes and sizes that fit on roads, and perhaps even more than that. PRTs tend to use uniform vehicles.
  • Robocars could trivially be privately owned, and express occupant personality. The wealthy can have a better robocar than the poor, and the wealthy will readily pay for that.
  • Robocars, by using all available pavement, can handle things like draining a sports arena after a game, a difficult challenge for PRT.
  • Robocars could use any different kind of fuel.

Bottom up vs. top-down

The real reason that I abandoned the PRT vision is that PRT is a top-down appraoch. Like all transit, it's specified, designed and bought by governmental transit planners, and purchased typically from a single supplier that goes through a government procurement process.

The robocar vision is a bottom-up one. The government probably certifies the safety of the vehicles, but they are bought, one at a time, by ordinary people. In particular, they are driven by the technophiles known as "early adopters."

Early adopters are rich and will buy technology ahead of the curve. They will waste money on it, and toss it in a year for the hot new version. The later adopters buy what the early adopters cast off. The early adopters drive all fresh start-up companies, and they demand constant innovation. And the entrepreneurs work to give it to them. The VCs invest in firms expecting those early adopters to be there.

This is how innovation works in computers and electronics. It's bottom up. Small innovators and individual buyers. Buying the latest things when it makes no financial sense. This pattern has been responsible for scores of amazing technological revolutions.

Robocars turn driving and transportation into a computer problem. They let Moore's "law" come to transportation. Every year, the abilities of the computers driving the cars get better and cheaper at the same time.

This is just not how municipal transit agencies work. They schedule in decades. They budget in hundreds of millions to billions. And in the computer world, what they spec out today will be obsolete long before they can get it deployed. They don't seem to be able to handle constantly changing their plans to deal with the new technology, and they've invested so much money into the old technology that they are scared to rethink. You get fired for rethinking, not for following through with the plan.

PRT advocates debate why they have never gotten a city to install a PRT. One reason is simply that city transit planners don't want to be the person who bet on a new technology. If it fails, that's a career ending move.

Government infrastructures are also rife with corruption and notorious in history for massive budget misestimation. We often see public infrastructure projects using well understood technology go 100% or 200% or more over budget.

PRT requires a monopoly on at least the track, either for the government itself or a franchised operator. While one can dream that many competing companies could operate independent cars on the common track, the reality is that today's PRT plans call for monopoly operation, with all the corruption and bad customer service that entails.

So while PRT systems might seem to have advantages, by the time a system planned today is in operation, I predict robocars will have driven circles around those advantages. We can't know this for sure, but it's the way computer technology works, when it is allowed to be driven from the bottom up.

For a real robocar revolution, the governments should do as little as they can. The monopoly is only on the blacktop, as it is today. The government will demand the robocars be safe, and require the occupants to obey any useful traffic regulations. (Many traffic regulations actually become unimportant with robocars. Speed limits and many other things don't matter -- all you really need are rules to be safe and not impede others unfairly.)

Governments can help a bit with broadcast of traffic information and light timing, and they can do more over time, but they should start simply, and only tweak where needed. Let the vehicles self-organize as they do today with humans behind the wheel -- but do it better than the humans do.

Pick the simplest platform

In the computer industry, we have learned that innovation happens fastest when you you standardize on the simplest possible platform, and let others innovate on top of it in many directions. The people of the future will know more than you about what they need and what is possible, so let them make as many of the decisions; don't make decisions for them.

This has been the winning strategy behind the internet, which is at its core extremely simple. It defeated competing network technologies where the network wanted to be "smart" and do things for you that it thought you needed. Turns out that this seems like a good idea when you first begin, but it eventually limits you to what was "smart" many years ago.

Plain flat roads are inferior by a number of metrics to "smarter" alternatives like tracks and guideways. But they allow any type of vehicle to run on them if they are wide enough, from bicycles to mining trucks. And most importantly, they allow what we haven't yet thought of. Indeed, since you can run rails down roads, they even allow the addition of railed vehicles, as they did with streetcars.

Energy experimentation

Privately owned vehicles on ordinary roads can't readily use power from (ugly) overhead lines and rails, but they can use every other kind of power, and more to the point, they can (viewed as a fleet) be using them all at once. This allows experimentation in energy storage and delivery, and the resulting innovation. Short-range lightweight vehicles already don't need very much battery, and as battery technology improves, they can benefit further from that.

Road vehicles can also try other things, like compressed air, compressed natural gas, biofuels, flywheels and of course plain old gasoline. While PRT pods could use such fuels, their central planning will make them slow to experiment and change, while early adopters will try out the new energy forms the moment they can buy them.

An all rail world

One of the biggest advantages of robocars over PRT or other track-based transit is the ability to work with existing infrastructure and private owner's money. Roads will be around, and maintained for many decades to come, so this money will already be spent.

PRT tracks/rails, it is argued, can be cheaper to build than roads. Perhaps in the very long term, we could see new urban regions only have PRT tracks, and minimal roads, it is argued. This could be a cheaper alternative. The problem is that PRT tracks are generally designed for lightweight PRT pods only. They are not capable of doing delivery of large and heavy items. So people will want some amount of access for trucks, and emergency vehicles. It's possible that could be done on cheap gravel roads, possibly with embedded PRT rails, but this seems unlikely. And then there are the bicycles.

(Heavy trucks, as it turns out, are the major cause of road wear. In both the PRT and robocar world, delivery of small packages is done by light vehicles which don't wear out the roads. Roads will be much cheaper to maintain.)

Rails are efficient but lack flexibility. One can't change lanes trivially or stop and park with complete choice of location, the way tire on pavement vehicles can. A stopped PRT pod blocks an entire section of rail, while an obstacle on the road can just be driven around.

Hybrid PRT, where vehicles have rail-wheels and tires, and can leave the PRT to go on roads (as HDVs or robocars) offers some possible solutions, but I'm hard pressed to see why, if you had a robocar ability, you would bother putting in the rails. The extra efficiency is minor compared to the construction cost and limitations those rails demand.

Isn't a guideway safer?

By modern standards, where we do not yet have the ability to build a commercial safe robocar, the natural instinct is to imagine that grade-separated guideways, as found in monorails, subways, people movers and elevated trains are the only safe way to have robotic vehicles. Today that's true. The goal of robocar technology is to eliminate the need for such guideways by creating systems that can work on ordinary roads, and deal with obstacles, pedestrians, animals, other vehicles and anything else that may alter the flow of traffic.

Guideways, especially grade-separated ones, seem safer but, since you also need roads (for trucks if nothing else,) even the cheapest guideways are hugely more expensive -- it's an entirely different ballpark -- and that is such a huge factor it controls whether they will exist or not, and how big the network can get. And even so, people find their ways into the guideways, either due to recklessness or because they wish suicide. Guideway based systems depend on the guideway for their protection -- if there is something on the tracks, they may very well not detect it, and they typically can't move out of the way if they do, they can only try to stop.

Suicide by robocar should actually be hard to do, once robocars can pass the school of fish test. I am sure that a determined suicide will be able to find a way to get hit by a robocar, but most people won't be able to. Suicide by guideway transit simply requires getting in the guideway today, even when there is a human driver.

Elevated guideways -- by far the most likely choice -- also must deal with the problem of evacuation, including evacuation for the disabled, young and infirm, in the event of problems ranging from vehicle breakdown to fire to earthquake. Waiting for a ladder truck may not suffice. In the case of a serious breakdown or accident that actually causes a vehicle to leave a guideway, elevated guideways add significant risk.

Robocar guideways

In the subset of places where guideways make a lot of sense, they can be built for use by robocars in a certain weight and size class. For example, they might accomodate only 1-2 person robocars, leaving larger and heavier vehicles to the roads below. Such guideways could scoot over high-congestion areas and intersections, and also into buildings.

Unlike typical elevated roadways, these would be narrow lightweight metal and concrete tracks with traction surface on the outsides and grille (to pass light) in the middle, perhaps only 6 feet wide. As such they would be vastly cheaper to build than bridges or elevated roads, requiring only a pillar every 100 to 150 feet for real estate, perhaps down the middle or sides of existing roads. They would offer traffic planners the ability to do something they can no longer easily do -- add road capacity.

On these guideways, with only robocars present, there would not be congestion. Nor would there be stop signs or traffic lights. As such they would add more capacity than a lane on the ground. Unlike PRT, however, they would only be built where there is high demand, and enough open space that the guideways are not visually disturbing. The guideways would have the opportunity to have spurs into the 2nd floors of buildings, taking passengers directly to elevators. The guideways could also offer an electrical charging line on the side for vehicles eqiupped with something akin to a trolley's pantograph.

As such, this could offer all the advantages of PRT while providing the universal coverage of robocars, with a much lower construction cost. This would also strongly encourage the use of small robocars by offering the ability to zoom over the most congested areas.

Robocars are still science fiction

It is a bold claim to say that a technology which is not yet fully here, and whose date of arrival can't be predicted, can make a more tangible (though still not deployed) technology be obsolete. In the strict sense, it can't.

Robocars are a Moore's law technology. The innovations from such technology have also never been sure, but it's never been wise to bet against them in the last few decades. In the next 5-10 years, we'll get a better understanding of how robocar technology is progressing. Is it zooming or is it stalled? How are governments around the world reacting?

So what I predict is that a time is coming soon when there is a reasonable chance that we'll know that PRT (and other transit) is obsolete. This puts a risk on building a large PRT (and on any other infrastructure based transportation technology.) And yes, there is a risk to not building one as well, but that's a risk socity has been willing to take for 30 years. So this is not a declaration of the complete obsolesence of PRT, but rather a warning that there are good odds this is coming. Nobody can guarantee you robocars will be here soon, but it may be unwise to bet that they won't. And a large investment in any other technology, in particular another untested technology is a bet that they won't.

Some PRT plans used PRT pods with tires that run on ordinary flat surfaces. Their guideways would be perfectly fine for robocars -- indeed the PRT pods are robocars. The world's firs small production PRT at Heathrow airport, is just such a system. A PRT system that builds such guideways is much less likely to become obsolete, though its pods will, over time.

Greater traffic capacity

While most PRT plans call for building or taking extra right-of-way, robocars should vasty increase the capacity of the existing road network, which turns out to be much larger than we might think. Consider the following:

  • Single person robocars, along with two-person face-to-face robocars, can use from 1/2 to 2/3 of the width of current lanes. Since most trips are one person, and most of the rest are two, this could almost double road capacity right there.
  • If needed, robcars can travel close behind other cars, much closer than humans can safely do. This can again double the capacity in a road full of robocars, and since robocars can find one another to convoy, it even works if only a fraction of cars are robocars.
  • During rush hours, on-street parking lanes can be quickly vacated by robocars which move out of the congestion zone, adding two full lanes (or 4 half-width robocar lanes) to most streets. A robotic tow-truck (with a human to hook up the chains, but who does not have to ride the truck back to the impound lot) could clear any HDVs blocking the lanes.
  • Robocars, timed to lights, should move at a quicker speed, and if not stopped by HDVs, should never collapse to "stop and go" traffic the way HDVs do today at rush hour. Stop and go has half the road capacity of smoothly flowing traffic -- yet another doubling of capacity. Robocars should be able to not need stop signs to be safe, further bumping capacity.
  • In a world where all cars have computer navigation, if not computer drivers, streets can be redirected. For example, in morning rush hour, 80% of the streets could become one-way towards downtown, and that could reverse in the afternoon. Of course these streets would also have no cars in their parking lanes at this time. (Some highways and bridges already are redirected every day by manual means.)
  • Robocars (and HDVs with advanced navigation) will listen for broadcasts about current traffic volumes on all city streets along their route, as well as predicted volumes for the future. They will automatically want to avoid highly congested routes and move to more empty routes. (If there is a reservation system, they may publish their intention to use a certain street at a certain future time, to allow other vehicles to decide they want to pick a different one.) As a result, the entire road grid can be fully used, for a vast increase in capacity. Cars will naturally gravitate to the roads with the lowest predicted utilization. Then as cars announce their intentions to use those roads their predicted utilization will rise, and the traffic can balance over all roads -- filling them with close to their safe free-flowing limit, but no more.
  • Finally, if traffic is very thick that this 20-fold increase in capacity (20-fold over the real-world congested, stop-and-go roads) is unable to handle it, we can have robo-buses, or trains using legacy rails. People wishing to travel a very heavily used route will have their robocar take them to an appointed place and time where they switch to a bus, which then travels nonstop along the route. At the end, it stops in a small lot where many single-person robotaxis are waiting to take them the rest of their way. This should allow a very large increase in the capacity of the roads, at the cost of having buses that are used only at these peak times -- still cheaper than building new right-of-way.


It need not be simply a choice between robocars and PRT. In fact, the current success story in the PRT field is ULTra, which makes a PRT that runs on rubber tires in simple paved guideways. ULTra's system wil open in mid 2010 as a parking lot shuttle at Heathrow Airport Terminal 5. Another tire-based PRT is being installed in Abu Dhabi's "Masdar City" -- a car free town being constructed there. (This system is currently on hold.)

Robocars and PRT, particularly tire-based PRT, have several problems in common, and solutions from one will migrate to the other, and act as proofs of concept.

Some of the problems in common include:

  • Basic smooth, stabilized riding along a chosen path, without the use of rails.
  • Navigation of a route, including reserving slots in the route, changing tracks and stopping at stations.
  • The slow parking operations at stations.
  • User interfaces for navigation and fare collection.
  • Security and handling of problems with the shared vehicles.
  • Vehicle provisioning and scheduling to handle demand.
  • Detection of obstacles and other vehicles, and slowing to avoid them. (PRT pods can rarely steer to avoid, but they can brake or accelerate to deal with problems.)
  • Logistic problems of building narrow paved guideways where they are needed.

Leave/Read Comments