Tag Archives: driverless cars

Short, clear station names vital to transit system wayfinding

WMATA map with long station names: “they’re not station names, they’re committee meeting minutes.”

The folks at London Reconnections have a new podcast – On Our Line. The second episode features a long conversation with two experts on transit map design and understanding, Max Roberts and Peter Lloyd.

The discussion hits on several topics about the challenges in transit map design, particularly for complicated networks. They also discuss objective measures of success in design (e.g. timing users in finding their way from point A to b on a map) and the conflicts with graphic design ideas. Another challenge is the future of the paper map and the seemingly inevitable move towards electronic map displays of some kind.

A few anecdotes stood out to me:

Touch Screen Maps: These might seem to be an obvious technological solution to mapping challenges with complex networks, frequent service changes, language barriers, etc. New York installed some touch screen maps as a part of a pilot program in 2014; despite rave reviews, no one seemed to use them. The podcast conversation (at 37:50) hits on the problems: the ad-supported model means the kiosks look like ads. Perhaps more interesting is the embarrassment of a rider using the kiosk, requiring a level of interaction that physically signals to everyone else on the platform that ‘I don’t know where I’m going.’ A static, printed map allows for consumption of information in a less obvious manner.

Station Names: Asked for examples of the worst transit maps they could think of, WMATA’s marathon-length station names are an obvious choice (at 1:07:20). Short station names are important to efficient, clear, and effective wayfinding. Roberts on WMATA’s map: “some of the stations – they’re not station names, they’re committee meeting minutes.”

File that one under “it’s funny because it’s true.”

Using the map to influence routing: Roberts obliquely mentions working with WMATA (48 minutes in) on changing the map to encourage different routing, presumably a reference to adjusting the map in order to encourage Blue Line riders from Virginia to transfer and use the Yellow Line (with excess capacity) to travel into DC.

It’s one thing for the map (or trip planner) to influence your route; it’s another for that decision to be made by an algorithm completely removed from human interaction. With driverless cars, it’s still unclear how humans will react to navigating networks in that way – adjusting human behavior is challenging enough.

I can’t believe I’m writing a post on Personal Rapid Transit!

Morgantown WV PRT System, as seen from Google Streetview

Reading through the history of the personal rapid transit (PRT) on the Verge by Adi Robertson, I couldn’t help but think of the similarities with many familiar projects. Cost overruns, scope creep, politics, government red tape, all conspiring to erode the value of an otherwise promising concept.

First, you can’t write about PRT without acknowledging the inherent geometric flaw of the concept: it can’t scale. Jarrett Walker frequently talks about the fundamental geometry of transit, and succinctly explains the geometric flaw of PRT:

Bottom line: When “personal rapid transit” succeeds, it succeeds by turning into a conventional fixed route transit system. The fantasy of “personal” transit is that a vehicle will be there just for our party and take us directly to our destination, but in constrained infrastructure this only works if demand is low. But PRT was meant to the the primary transport system in a car-free city, so demand would be high. It was never going to work.

This is also true of the Morgantown, WV PRT system, which makes use of different operating modes. During times of high demand, it operates as a fixed route transit system between the busiest stations; during low demand periods, cars stop at every station, regardless of demand.

Mass transit might be an out of fashion descriptor, but it helps illustrate transit’s scalability. Good transit doesn’t just move large masses of people, it requires mass to succeed. ‘Personal’ transit rejects the masses; it also requires expensive infrastructure to inefficiently move people.

Robertson skirts around the geometric limitations of PRT as a concept, but never appropriately douses the concept with cold water. Any history of PRT must focus on the Morgantown, WV system. Any article about PRT will inevitably draw comparisons to current research on driverless cars. Comparing the two exposes the conceptual flaw:

Self-driving vehicles, he points out, wouldn’t have taken cars off Morgantown’s crowded roads — at least, not in the same volume. As long as they’re intermingled with human-driven cars, they can’t run with the same centralized efficiency. And once you start thinking about the obvious solution — a dedicated lane for self-driving cars — you might start running into the same problems as PRT.

Leaving aside PRT’s conceptual flaws, Robertson’s history of the concept echoes common challenges in the American history of infrastructure projects: shifting government mandates, political interference, procurement regulations, and so on. Some highlights:

Goals for transit: Robertson documents the history of federal funding for PRT, with the Urban Mass Transit Administration providing research grants to explore the concept.

The focus on new technology in transit often meant unnecessarily reinventing the wheel (see BART’s broad gauge track), but also exploring new concepts like PRT. New concepts are sexy, even attracting the direct interests of President Nixon:

His mantra, as Alden puts it, was that if “Kennedy can get a man on the Moon, I can get a man across Manhattan.”

Lack of clarity about the UMTA’s goals for the program help add to the confusion. Is the goal to provide effective transit, or to prove a new technology/concept? Crosstown transit is a practical goal, but it doesn’t require big technological innovations. Landing on the Moon is an impractical goal that wasn’t possible without new technologies – and the moonshot analogy makes it easy to conflate two different goals.

From the start, there’s tension between researching new technologies and practical, proven, cost-effective projects. Many PRT boosters in West Virginia were approaching this a big experiment; the government bureaucrats wanted a functioning system. Once the system proved more conventional than revolutionary, Robertson notes, “the age of experimentation was over.”

Politics: Robertson also shows the competing interests of the various parties involved in funding and executing the Morgantown project. West Virginia University approached PRT as an experiment, while UMTA wanted a more practical proof of concept – something that could be built elsewhere if successful. On top of these turf battles, President Nixon wanted a completed project to include in his re-election campaign materials, pressuring the team to complete things before they were ready.

Procurement and red tape: As WVU championed the PRT project, they looked for federal funds to offset the cost. Then, as now, those dollars had strings attached. UMTA required a NASA JPL redesign of the vehicles; one of the independent engineers took patents to established defense contractor Boeing in order to better compete in project bidding.

Right of way: The single most important element of the Morgantown PRT system is the elevated guideway. Complete grade separation from the traffic at street level and the interference from cars, bikes, and pedestrians not only speeds travel, but made PRT’s automated operation possible (note: this remains true, it should be far easier to automate a subway system than to create a fleet of driverless cars).

Despite the inherent geometric challenges of personal transit as a service, the system nevertheless demonstrated the value of guideways; and also the reasons why we don’t have more of them: local opposition and cost. One PRT booster:

To Kornhauser, the issue is less that the technology was inherently inadequate than that it was expensive and inconvenient. “You didn’t need that much intelligence in the vehicle to be able to do all this stuff,” he says. “The problem was that nobody really wanted to invest the money to build the exclusive guideway. That’s the short and the long of it.”

And Robertson on the local opposition to erecting concrete guideways all over the city:

Even the most time-tested (and desperately needed) public transit systems have trouble securing space and laying track; New York City’s history is littered with unbuilt subway lines that were killed by local protests and a lack of money. PRT guideways had some advantages over trains, like their near-silence, but they would still require cities to build miles of concrete chutes. And unlike a subway line extension, there would be no guarantee that people would accept the new system. Or, as one former transportation commissioner told NPR when asked about personal rapid transit last year: “The last thing you want to do is put up some track all over the place and have it just there.”

Also, unlike a more traditional elevated line (something I’ve defended here previously), the ideal of PRT means offering door to door transit, which in turn requires a guideway of some kind from door to door.

Driverless cars: implications for city planning and urban transportation

Nevada autonomous vehicle license plate. CC image from National Museum of American History.

Building on the implications of driverless cars on car ownership, as well as the notion that planners aren’t preparing for the rise of autonomous vehicles, I wanted to dive further into potential implications of widespread adoption of the technology. Nat Bottigheimer in Greater Greater Washington argues that city planning as a profession is unprepared for autonomous vehicles:

Self-driving cars address many of the safety and travel efficiency objections that Smart Growth advocates often make about road expansion, or the use of limited street space.

Part of Bottingheimer’s concern is a lack of quantitative analysis, particularly as it relates to the impacts of self-driving cars. However, the real debate is about qualitative values that feed into our analysis.

The officials responsible for parking lot and garage building, transit system growth, bike lane construction, intersection expansions, sidewalk improvements, and road widenings need to analyze quantitatively how self-driving cars could affect their plans, and to prepare alternatives in case things change.

There is one over-arching problem with this approach: our current quantitative analysis all too often is nothing but bad pseudo-science. Donald Shoup has extensively documented the problems with minimum parking requirements in zoning codes, for example. Here, poor policy with vast unintended consequences is based on some level of flawed quantitative analysis, the kind that does not acknowledge the inherent uncertainty in our understanding or ability to project the future. Instead, the analysis is based on assumptions, yet the assumptions are really value-laden statements that carry a great deal of weight.

Even the very structure of the planning and regulation for the future carries a bias: a requirement to provide parking spaces in anticipation of future demand will, by nature, ignore the complexity of the marketplace for off-street parking and the natural range of parking demand.

Bottigheimer is also concerned about the impacts of self-driving cars on future land use forecasts:

Planners need to examine how travel forecasting tools that are based on current patterns of car ownership and use will need to change to adapt to new statistical relationships between population, car ownership, trip-making, car-sharing, and travel patterns.

By all means, we need to adjust our forecasting tools. However, we shouldn’t be doing so simply based on the arrival of a new technology. We should adjust them because they’re not particularly accurate and their erroneous projections have large impacts on how we plan. Driverless cars aren’t the problem here. The problem is in our assumptions, our inaccurate analysis, and our decision-making processes that rely on such erroneous projections.

Leaving the limitations of quantitative analysis aside for the moment, we can still hypothesize (qualitatively, perhaps) about the future world of driverless cars. Assuming that autonomous vehicles do indeed reduce car ownership and begin to serve as robo-taxis, we can sketch out plausible scenarios for the future. We assume car ownership will decrease, but vehicle-miles traveled may increase.

City Planning and Street Design:

One of Bottigheimer’s chief concerns is that “planners and placemaking advocates will need to step up their game” given the potential benefits for safety, increased car capacity,

As mentioned above, much of the ‘safety’ benefits are about cars operating in car-only environments (e.g. highways), when the real safety challenges are in streets with mixed traffic: pedestrians, bikes, cars, and buses all sharing the same space. In this case, the values planners and placemaking advocates are pushing for remain the same, regardless of who – or what – is driving the cars. The laws of physics won’t change; providing a safe environment for pedestrians will still be based on the lowest common denominator for safe speeds, etc.

The biggest concern should be in the environments that aren’t highways, yet aren’t city streets, either. Will driverless cars forever push stroads into highway territory? Borrowing Jarrett Walker’s phrasing, technology can’t change geometry, except in some cases at the margins.

Instead of a technical pursuit of maximum vehicle throughput (informed by quantitative analysis), the real question is one of values. The values that inform planning for a place or a street will set the tone for the quantitative analysis that follows. Maximizing vehicle throughput is not a neutral, analytical goal.

Congestion:

Congestion is a more interesting case, as it will still be an economic problem – centralized control might help mitigate some traffic issues, but it doesn’t solve the fundamental economic conundrum of congestion. Here, too, the economic solutions in a world of human-driven cars will have the same framework as one with computers behind the wheel.

Driverless cars might change the exact price points, but they don’t alter the basic logic behind congestion-mitigation measures like a cordon charge in London or Stockholm, or like Uber’s surge pricing (efficient and rational as it might be, but perhaps too honest). Again, technology can’t fundamentally change geometry. Cars will still be cars, and even if driverless cars improve on the current capacity limitations of highways, they do not eliminate such constraints.

Qualitative Concerns:

Instead of twisting ourselves in knots over projections about the future that are sure to be wrong, planning for autonomous cars should instead focus on the values and the kind of places we want to plan for. We should adjust our policies to embrace the values of the communities (which alone is a challenging process). We should be aware about the poor accuracy of forecasts and work to build policies with the flexibility to adapt.

Driverless cars: a city of cheap robotaxis and the end of car ownership

CC image from the Museum of American History.

To date, most of the writing about driverless cars seems to focus on technology’s potential to make driving safer by eliminating collisions between vehicles. The thinking is similar to other auto safety improvements such as air bags or anti-lock brakes. These technological advances (endorsed by the US DOT) incrementally improve the safety of those driving – assuming that you are using a narrowly focused definition of ‘safety.’ However, an auto-centric definition of safety only works in auto-centric environments; in urban environments where cars and bikes and pedestrians are all sharing the same space, the definition of safety cannot solely focus on eliminating collisions between high-tech cars (more on this later).

Other articles predict that driverless cars mean the end of transit – an unlikely scenario that ignores the basic geometry of car-based systems and the capacity advantages of transit (imagine shutting down New York’s transit system and trying to fill that role with nothing but taxis – good luck). Furthermore, if driverless cars make vehicle automation easy, then it should also help drive down the costs for automating transit itself (among other potential uses) and unlock the benefits of automated transit.

Ownership:

The far more interesting scenario is one where autonomous vehicles completely upset the benefits of owning your own car. In the Atlantic Cities, Eric Jaffe questions the assumptions of car ownership in a world of driverless cars:

But we’re not so far away from this future that it’s too early to start considering what it might look like. As Matt Yglesias wrote at Slate in August, Google, the leaders in autonomous car technology, must have had some vision in mind to shell out $258 million for the car-slash-ridesharing service Uber: “ubiquitous taxis — summoned via smartphone or weird glasses — that are so cheap they make car ownership obsolete.”

Think about this world of shared autonomous vehicles for a moment. You wake up and get ready for work, and a few minutes before it’s time to leave you press a button and order an SAV [Shared Autonomous Vehicle]. The car has been strategically positioned to wait in high-demand areas, so you don’t have to wait long. You might share the ride with a couple travelers just as you share an elevator, or perhaps pay a premium to ride alone. Either way, you clear your inbox or read the paper during the commute, which is safer and more reliable than it used to be.

So, basically Robo-Uber. Or Auto-Car2go. Or Johnny Cab. This kind of behavior seems to be a far more likely outcome of the technology than the continued paradigm of each individual owning a car for personal use. Just as transit consultant Jarrett Walker talks about the importance of frequent transit service in providing freedom for users, the on-demand nature of the personal car is similarly freeing – but it required a) ownership of the car to ensure on-demand use, and b) the owner to actually do the driving.

Travel Behavior:

But what kind of changes in behavior can we expect from this shift away from car ownership? Writing at Greater Greater Washington, Nat Bottigheimer notes that planners haven’t even begun to address the issue. Jaffe’s article, however, cites some preliminary research from Austin on the impact of robotaxis.

Civil engineer Kara M. Kockelman of the University of Texas at Austin recently modeled the potential ownership change with grad student Daniel Fagnant…

The results offer an enticing glimpse of a world without car-ownership. Each SAV in the Austin model replaced about 11 conventional household vehicles. The roughly 20,000 people who made up this shared network, formerly owners of roughly as many cars, were now served by a mere 1,700 SAVs. Travelers waited an average of only 20 seconds for their ride to arrive, and you could literally count the number who waited more than 10 minutes on one hand (three). That’s to say nothing of personal savings in terms of cost (insurance, parking, gas) and time.

“Even when we doubled or quadrupled or halved or quartered that trip-making, we didn’t have big changes in our key variables,” says Kockelman. “This replacement rate, this eleven-to-one, those things were very stable.”

Kockelman is quick to point out the caveats. The biggest is that for all the savings in private car-ownership, vehicle-miles traveled doesn’t go down in the Austin model. In fact, it goes up about 10 percent. That’s because not only are SAVs making all the trips people used to make on their own, but they’re repositioning themselves in between trips to reduce wait times (see below). The additional wear also means manufacturers produce about the same number of cars, too, though each new fleet is no doubt a bit smaller and cleaner than the last.

So, a huge decrease in the total number of cars (presumably, with a corresponding decrease in parking demand, making the already-questionable logic behind zoning code parking requirements even more dubious) but an increase in the total vehicle miles traveled indicates that such technology won’t be a magic cure for congestion. It won’t spell the end of public transit in our cities. If the safety benefits accrue mostly to highway travel, it won’t change the need for safer streets where pedestrians, bikes, and cars mix.

The next question is on the impacts of driverless cars on cities and city planning.

What would change with driverless cars?

Robocar electronics - CC image from Steve Jurvetson

If we can agree that technology doesn’t change geometry, and therefore driverless cars won’t substantially change the fundamental capacity and spatial requirements of our current auto-based transportation systems, then what would they change?

Chris Bradford takes a stab at this question, taking note of Matt Ygleisas’s prediction of reduced demand for parking. Matt cites the idea of having a driverless car drop you off at a commuter rail station in the morning in order to make use of the higher capacity rail system to enter the city (thanks to the relevant geometries of rapid transit), while the car would then return to your house – eliminating the need for more car storage at the rail station. Chris takes that one step further, noting that with a tireless ‘driver,’ the needs for vehicle storage wouldn’t need to use a set space at all, but could be accomplished through cruising.

While both ideas would reduce the need for parking spaces, they would also increase the VMT for any given trip – either through cruising for parking or for increased deadhead trips, further clogging the streets. This might not be a problem in certain cases where congestion isn’t currently an issue, but it sure wouldn’t help in places where congestion is already a problem. Bradford notes this:

In fact, this perfectly rational practice will probably be so harmful, so patently selfish, so despised that it will be necessary to outlaw it. Which means everyone will still have to find a spot for his car, driverless or not. Which means that, despite the title of this post, we might not see a robocar apocalypse after all, or a parking bubble, either (other than the existing bubble that local governments have created with underpriced street parking and mandatory parking minimums.)

Perhaps the most interesting application, then, isn’t the need to store a car for personal use (given the issues of storage raised above), but to allow that car to be used productively by someone else. A driverless taxi, otherwise (hence my choice for my previous post’s image of Total Recall’s Johnny Cab – I don’t know if the new version of the film this summer will depict the Johnny Cab, if it does so at all).

You can already see the convergence of different car ownership models. A taxi is owned by an operator, they provides rides for hire, charging you for the convenience of the trip in their car and for not having to drive yourself. Compare that to the current point-to-point carsharing model like Car2Go, and the only real difference is the driver. Both charge based on time and/or distance traveled, both offer point to point trips in a vehicle you don’t own.

While the cost of these robocars would likely come down over time, they’d still be more expensive than regular ol’ human-driven cars, meaning that the trends towards collaborative consumption would continue, and the robocars would serve their best use as taxis. The value of owning one yourself would be limited, unless you had a ton of disposable income.

As Matt Yglesias put it, “imagine a world of cheap, ubiquitous taxis.” The net impact would be favorable to cities and those who live in them. The limits of the automotive geometry and capacity wouldn’t fundamentally change, so this would still be a premium service over much higher capacity mass rapid transportation. The benefits of owning a car would continue to decline in urban areas, as would the cost of the auto-based alternatives (like taxis).

Driverless cars don’t change geometry

Via the Streetsblog Network, I came across this Salon piece from Michael Lind praising our future driverless car overlords. Angie Schmidt at Streetsblog did a nice job to take down some of Lind’s loaded language, particularly the bits about “rigging markets” (which rings just as hollow as the cries about “social engineering” – as Timothy Lee notes, there’s no such thing as an intervention-free infrastructure policy).

Those issues aside, the biggest thing that Lind misses isn’t about technology at all – but rather about geometry, land use, and the relationship between transportation and the built environment. Lind writes:

As the white windmills fade from the picture of the future, so do the bullet trains speeding past them. Even before the end of President Obama’s first four years, unrealistic fantasies about high-speed passenger rail had collapsed. Federal funding for high-speed rail demonstration projects has been minuscule and symbolic. State and local governments continue to conclude that the costs of high-speed passenger rail outweigh the alleged benefits.

In the longer run, robocars may be fatal for fixed-rail transportation, at least for passengers rather than freight. Google has been test driving self-driving cars in California and Nevada has become the first state to legalize driverless vehicles. No doubt it will take several decades for safety issues and legal arrangements to be worked out. But high-speed trains might find competition in high-speed convoys of robot cars on smart highways, allowed higher speeds once human error has been eliminated. And the price advantage of subway tickets over taxi fares in cities may vanish, when the taxis drive themselves. Point-to-point travel, within cities or between them, is inherently more convenient than train or subway journeys which require changing modes of transit in the course of a journey. Thanks to robocars, much cheaper point-to-point travel everywhere may eventually be cheap enough to relegate light rail and inter-city rail to the museum, along with the horse-drawn omnibus and the trans-atlantic blimp.

Paraphrasing Jarrett Walker (aside: his recently published book is an excellent read), technology does not change geometry. A driverless car is still a car, the geometry that governs the car is the same regardless of who (or what) is at the controls. Despite predictions about how this technology could change everything (see a whole series of GGW posts), I find the possibility for change to be marginal. Driverless Johnny Cabs, Total Recall-style might decrease the cost of providing taxi service, but that won’t fundamentally change the inherent capacity limitations of taxis compared against a subway system.

The choice of the taxi as a demonstration for the technology is interesting. Most taxis operate in big cities, and big cities tend to be dense. Density helps support high levels of transit service and ensures that lots of potential trip destinations are easily reached by foot or by transit, thereby diminishing the market for these automated taxis. Cars, regardless of who’s driving, don’t have an advantage in point to point travel over pedestrians, transit, or other modes in cities.

The other point Lind makes is in investment priorities for government-funded infrastructure (hence the earlier comment about “rigging markets”). Lind seems to view the built environment as static, rather than an evolving system that changes in concordance with the changes to the transportation infrastructure. New York’s subways fueled its dense development, and that density in turn provides the market for high capacity rapid transit. Given growing populations and constantly changing cityscapes, these infrastructure investments in transit are step along the process of letting out cities continue to grow.

(semi-related sidebar on growth patterns: check out this article in Scientific American on the patterns of growth among subway networks around the world. The authors concluded ” that the geometries of large subway networks are guided by simple, universal rules.” – reminiscent of Geoffrey West et al)

City Block

a work in progress…