Happy 60th Birthday, Interstate Highway System! You Look Awful | Wired

Aarian Marshall at Wired writes “Happy 60th Birthday, Interstate Highway System! You Look Awful.” The author discusses my work with Matt Kahn:

In 2011, transportation economists Matthew Kahn and David Levinson laid out a kind of radical idea—what if the country stopped focusing on building new highway stuff, and got down to fixing what it already has? In the long run, this would save some serious dough: For every $1 in gas tax revenue spent on your maintenance, the government would save between $4 and $10 on future repairs.

Sometimes a city is a tree

Christopher Alexander wrote a brilliant essay in 1965: “A City is not a Tree“. Long interested in Alexander’s work since graduate school in Berkeley, I recently re-read the piece which has been packaged in a 50th anniversary volume. The original article is available online free. There have been numerous other papers that have commented on various aspects Alexander’s piece, I can’t list them all, it has been cited over 1000 times.


Alexander criticizes new towns, notably my home town, Columbia, Maryland, for being treelike in its conception, rather than what he terms a semi-lattice, but we would more informally call a mesh-like network. The neighborhoods belong to villages, the villages are part of the city. The neighborhoods, following an element of the Radburn plan, are isolated, that is, one cannot go from one neighborhood to another without being on a village street. That’s not to say there is no through traffic, there is, because the neighborhood network has more than one outlet.

But it’s not simply the street network that is tree like, retail is also tree-like. The neighborhood center might have a convenience store (7-11 or Wawa), the village center would have a grocery and 10-20 smaller stores. Town Center had the Mall in Columbia with department stores and over 100 shops. Perhaps Walter Christaller would be proud that Central Place Theory was not merely descriptive, but also prescriptive.

Finally the schools were tree-like. The neighborhood elementary school fed the village-level Middle and High schools.

In practice it was not so tree-like. As a resident of the neighborhood of Bryant Woods in the Village of Wilde Lake, my mom could go shopping at a supermarket at Joseph Square shopping center in the Village of Harper’s Choice. It was only another mile down the road. Later as a resident of Longfellow neighborhood, I could open-enroll into Wilde Lake Middle School. And of course not every neighborhood got an elementary school, not every village got a Middle or High School, and the boundaries were overlapping. So while the stylized schematic drawing may have looked treelike from the perspective of an architect a few thousand miles away, it was not treelike in practice on the ground. See The Next America Revisited for my take.

treenessI don’t believe the planners envisioned it would be so tidy — though better to start out tidy and let entropy emerge rather than start out chaotic and hopes it self-organizes into an aesthetically pleasing environment. Rather my impression from reading a lot of the documents and having lived there and hearing talks from Rouse company officials and so on is that they believed that treelike street networks reduced through traffic, just as planners favor traffic calming today. They believed shopping should be organized into centers, rather than sprawled out uncontrolled along streets, and they should be spaced to be closer to residents. They believed local government is somewhat hierarchical (national, state, county, city) and that village was an organization unit that had some value to regulate things at a more local level than the city (Columbia is not technically a city, it’s just a home owners association, though it is a Census-Defined Place and the second largest in Maryland, after Baltimore, just as was foretold in the 1960s). They believed kids should walk to their neighborhood school, so the neighborhood should be the right size to support the school, which should have X students for pedagogical and cost-efficiency reasons, and ideally students would walk to their middle and high school too, but that middle and high schools should be larger. So the hierarchy was a natural way of organizing that.

But even if Columbia is innocent of being as treelike as Alexander feared (and certainly some new towns were more treelike), the suburbs are certainly more treelike than cities. My students have measured the “treeness” of networks, introducing the metric in Xie and Levinson (2007) Measuring the Structure of Road Networks. For instance in Network structure and the journey to work: An intra-metropolitan analysis (under review) by Pavithra Parthasarathi and myself, we see that treeness is not surprisingly higher at the suburban edges of the metropolitan area than in the center, though it declines as we see rural areas, where the sparser network is also more mesh or grid-like. (See figure)

Still, sometimes the city is a tree, or at least aspects of it are. In particular, many networked physical infrastructures are better organized as trees, especially if they require a large capital investment (like a waste water treatment facility). Similarly, the stream and river valleys are naturally organized as hierarchies. Transit networks are also often more treelike or radial than roads, and while may eventually evolve into ring-radial system, don’t generally start out that way. See Roth et al. 2012.

This transit network looks pretty tree-like.

Clearly social connections should not be assumed to flow in a way that maps directly to the physical layout of the network, all other things being equal, you are more likely to know your neighbor than a randomly selected person farther away. Yet, in a modern world with migration and telecommunication, you are likely to know someone specifically who is not on your block and to not  know everyone on your street. Growing up, my mom’s friends were scattered across Columbia, not just in her neighborhood. ‘Community without propinquity’ was first identified by Berkeley Planner Mel Webber, and certainly applied in Columbia as it does everywhere, where people could meet based on any kind of interest, not simply the desire to live on the same street. The physical form of the city does not represent how the city works, but more importantly the plans do not determine how the city works. People and their relationships are affected by their environment, and reshape it to suit their needs.

Initial impressions of A Line bus mostly positive | Star Tribune


Janet Moore of the Strib writes: “Initial impressions of A Line bus mostly positive.” I get a quote:

University of Minnesota Prof. David Levinson recently released an accessibility evaluation of the A Line, part of a broader, federally funded project. The study found that workers living within a half mile (or a 10-minute walk) of the A Line can reach 11 percent or 4,500 more jobs within 30 minutes of the new service than before. Employers in the same area can reach 6.4 percent more workers, or 2,000 additional employees.

“We found the net accessibility [of the A Line] is positive overall, so more people are winners than losers and the losers don’t lose very much,” Levinson said.

Roth says the A Line appears to be attracting new riders who like the new stations and the light-rail-like amenities over the local bus.

Levinson noted: “You won’t change your behavior on Day One, it takes awhile to build. Over time, we expect people to use the A Line more than the 84. And if more people use it, then it justifies the investment.”

Note the link to our report is added. The Strib surely meant to link back to the original report, but couldn’t because HTML is hard.

(Update: 6/28 Now linked to)

The Hierarchy of Roads: 7 Axioms on street design

I posit several Axioms about the hierarchy of roads.

Axiom 1: Some roads should be fast

The aim of transport is connecting people with destinations. They can connect with more destinations if they can do so in less time. Ceteris paribus, faster roads will take less time. Without loss of generality, let’s call these roads highways.

Axiom 2: Some roads should be slow

Some roads serve neighborhoods and have traffic that is not just motor vehicles. Ceteris paribus, slower roads are more likely to ensure safety [both reducing the probability of a collision through higher reaction times and reducing the impact of a collision should one occur], a high quality of life, and increased interaction within the neighborhood. Without loss of generality, let’s call these roads streets.

Axiom 3: Fast roads (highways) attract traffic from slow roads (streets)

In general, people prefer to spend less time traveling, and will spend less time on faster roads. These roads will attract more people. There will be net reductions in traffic on streets that are made slower and net increases in traffic on roads that are made faster.

Axiom 4: Urban design, congestion, safety, and funding problems arise when streets and highways are confused.

People, who are soft and move slowly, do not mix with vehicles, which are hard, when they move fast. If people feel unsafe they will avoid the place. Streets functioning as highways and managed by higher levels of government will be redesigned to be highways, — what Charles Marohn of Strong Towns calls “stroads” — destroying their street function.

Streets types matrix from Transport for London looking at tradeoff of Movement and Place. http://content.tfl.gov.uk/street-types-matrix.pdf
Streets types matrix from Transport for London looking at tradeoff of Movement and Place.

Further trying to move highway levels of through traffic on roads initially designed as streets with lots of access and at grade intersections is a natural misfit that will result in local congestion. At least this limits the amount of through traffic. Traffic and demand comprise a negative feedback system, more traffic slows speeds –> slower speed lowers demand –> less demand reduces traffic.

Axiom 5: Without strict controls, properties will try to gain direct access to highways.

Many streets started out as highways in previous generations with earlier technologies.  They were once crossroads that attracted businesses and became a place. This is the dual or mirror of the “Stroads” problem, in analogy, we might call them “Reets“.

While this origin story is not of itself a problem, the road should be designed for what it does, and what we want it to do, not what it once did. Highways with traffic are attractive places to open businesses. The US Highway System (the national system before the Interstate, which still exists) was plagued with this problem, once freeflowing roads were subject to steady speed deterioration as new motels, gas stations, restaurants, and stands emerged to exploit the traffic. By design to overcome this problem, the Interstate was more regulated in this regard, and was instead a limited access facility.


Axiom 6: Successful streets will attract more traffic.

Streets that have lots of local activity will encourage vehicle traffic as people seek to take advantage of the activity, and park their vehicle nearby. This does not justify “upgrading” the street through widening, which takes out the very elements that made it successful in the first place.

Axiom 7: The Hierarchy of Roads is an emergent process.

Even in the absence of central planning, a hierarchy of roads would emerge. Some roads will become more important than others just because of randomness, geography and topology, and positive feedback effects. Local roads naturally serve more local traffic, and we can distinguish the importance of roads by the source of their users. While A City is Not a Tree as Christopher Alexander said, it does have hierarchical features.

New automated vehicles can be better regulated than mere humans. There will also be a new Cambrian explosion of vehicle forms which are specialized for markets, especially in urban areas where mobility as a service is plausible. This is a huge infrastructure opportunity. We  should redesign our road hierarchy with these axioms and the possibility of slow vehicles becoming mainstream or at least standard. We should think about developing an interconnected slow vehicle network so that small neighborhood vehicles (think souped up golf carts) cannot not only travel within neighborhoods or on campuses, but between adjacent neighborhoods, without attracting longer distance traffic, where slow and fast vehicles need not mix.

There should be interesting designs for this, which are not today’s standard recipes, since this is as much at the level of network design rather than road design.

Elements of Access: Unitary vs. Modular Architectures

Designs might be comprised of individual, discrete building blocks (modules) that are combined into a pattern, or may be holistic (unitary) so that a small part cannot simply be interchanged with something similar without breaking the whole design.

Most things are combinations of the two. Software has moved very much to modular architecture, and as systems become large and complex, this is a logical way of reducing complexity. On the other hand, there are advantages of integration, an airframe integrating the fuselage with a wing might be more aerodynamic or less expensive to construct or more structurally sound than a fuselage with wings attached (I don’t know any of these things, but they might).

Another brick in the road.
Another brick in the road.

In surface transport we have lots of modules: vehicles and infrastructure are often separated (in elevators they are not), bridges and roads are distinct, each link is a separate module, but you can’t build half a link and expect it function. Poured asphalt is more unitary than individual bricks.

Traffic engineers operate on a system designed by highway engineers and planners and consider, for instance, the traffic signal timings as a distinct element that can optimized with everything else (lane configurations, pavement, etc.) fixed. Traffic engineers recognize that traffic signal timings affect the quality of flow upstream and downstream, and so will often time signals as a system, to optimize flow not just at the signal but for a corridor or a city network. This is recognition of a unitary aspect of the road network.

However this unitary nature of the network logic breaks the unitary nature of a neighborhood, where we might want the signals configured for pedestrians, and we might want roads redesigned to serve local rather than regional needs.

Elements of Access: Transport Planning for Engineers, Transport Engineering for Planners. By David M. Levinson, Wes Marshall, Kay Axhausen.
Elements of Access: Transport Planning for Engineers, Transport Engineering for Planners. By David M. Levinson, Wes Marshall, Kay Axhausen.

A modular architecture, where the signal is timed independently, obviously can do no better than the systemwide performance metric (mobility) of the overall system, but at least enables the maximization of the quality of the neighborhood if the appropriate local settings are chosen for the traffic signal module (at the expense of system-wide optimality on a mobility dimension). Similarly a purely local design may reek havoc with systemwide flow and have implications elsewhere on the network. While a module can only optimize for one master (and potentially less optimally than a unitary design), it can alternatively satisfice across multiple masters. In contradistinction, a unitary architecture must sacrifice one master for the sake of another. Modularization provides flexibility at the cost of at least one dimension of optimality. A unitary design lacks flexibility and adaptability. A new wing design will not help a unitary airframe.

We might further think about how systems change, either piece by piece or by replacement of the whole. Is a ship where every plank has been replaced one-by-one over the course of time still the same ship? We usually conclude it is. Is a ship which is burned down and replaced with a replica the same. We conclude it isn’t. We will rename it, or at least give it a number or letter to distinguish it (Enterprise D). Continuity is important, and is enhanced by modularity.

The job of the designer is to understand these tradeoffs and select appropriate architectural strategies (unitary vs. modularity for particular design choices), and then design the modules as appropriate. While there is no one true path (this is not religious), there are consequences and values at play.

Sci-fi style transportation is coming to a city near you | Metro

I was interviewed by Daniel Casillas of Metro newspaper for an article “Sci-fi style transportation is coming to a city near you“. Article selection below:


“The major new technology change is the introduction of autonomous vehicles (AVs), first on high-end cars, and then through all new vehicles, before eventually replacing all cars. This will greatly increase safety, and as deployment becomes widespread, road capacity will be increased,” explained David Levinson, author of the book “The End of Traffic and the Future of Transport.”

But how can we get faster and more efficient transport systems? According to the experts, we need a combination of broad computer control and a move from fossil fuels to electricity. These advancements would largely accomplish both goals, particularly if the remaining oil and coal-fired generating plants were replaced by cleaner and cheaper alternatives. This combination would make personal transportation faster, greener and cheaper.


Metro sat down with David Levinson, author of the book “The End of Traffic and the Future of Transport,” to discuss his views on the new age of transportation.

Do you see flying cars as a real option of transportation in the near future?

No, not in the near future; I think we are decades away from widespread use. While in some ways flying cars are simpler (there is more space up there than down here, so there’s a reduced likelihood of crashing), in other ways, they are more complex, and ensuring safety and stability will require a lot of proof. The energy requirements are probably also greater, as takeoff is energy intensive.

What are the main challenges for the new ways of transportation?

Getting the technology right is the main challenge. While already automated vehicles (AVs) are safer than human drivers, there is still uncertainty. Ensuring that sensors are reliable enough and the algorithms are good enough that humans don’t need to pay attention at all is the critical turning point. The costs will come down with mass production of the sensors, and if we can remove the steering wheel and brakes from human control, we can make the car both less expensive and safer. But deployment is a gradual process. It won’t happen all at once.

How can we get faster and more efficient transport systems?

Most cities in the world do not price their roads, they are allocated as first-come, first-served. Thus we get congestion when we underprice roads. For most other goods, we charge more in peak times (think about airlines, hotels, restaurants, movie theaters, and even public transport). A few cities like London, Stockholm, and Singapore are experimenting with road pricing, and this is an important solution to the congestion problem to better balance loads across the network.

Do you see the popularization of cleaner fuels among new transport? 

I expect electric vehicles to become more common over the coming decades. We are already seeing some European countries implementing a phase out of the internal combustion engine and gasoline for environmental reasons, but we also need to keep in mind that electric vehicles are simpler and less expensive than traditional gasoline-powered cars. The main drawback has been the large battery requirements in terms of space and the cost of batteries, but energy efficiency from batteries has steadily been getting better and charging stations are more widespread. The popularity of the new Tesla, with hundreds of thousands of pre-orders for a vehicle more than a year away, indicates the popularity of such cars once they become affordable

—Daniel Casillas 

9 Apple software problems

I first wrote this six months ago, but it sat in my drafts bin. I have updated and posted it as a record of issues with Apple software.

There is a classic tradeoff between economies of scale, which bigger organizations can exploit to lower costs, and span of control, which makes it harder for management to keep the eye on the ball. I am a long-time Apple User (Since 1981), and have gone through ups and downs. I am not keen on switching platforms or ecosystems, and of course, like everyone, am somewhat sucked into Google-world if only because of work. But why are these problems:

  1. Mail, after moving a message to Archive, Mail quickly moves it back to the inbox for a brief time before moving it back to Archive. This might have something to do with Gmail, but this behavior did not used to appear.
  2. Buying an app for a kid (subject to restrictions so they cannot buy their own apps) in my family on their iPad (an old hand-me-down without TouchID) requires the same password be entered accurately at least 4 times. Sometimes TouchID doesn’t work on parent’s iPhone to approve, instead another password must be entered. I would think 2 passwords (tops) should be sufficient. I suspect this is a problem with Apple ID services not talking to each other, so that there is one required for buying and one required for asking parents and then it forgets and then it forgets again? If I remember I instead buy on my device and it shows up on theirs using only TouchID. But this is not natural
  3. There are no parental controls on apps on the Apple TV (e.g. blocking YouTube). Why can’t I switch states to separate adults from kids, or who is watching on a shared device. My kids should have access to different GameCenter statistics than I do.
  4. iCloud Storage isn’t shared across the Family. If I have extra iCloud storage, family members should be able to use. Music has a family plan, why not storage?
  5. Music on “Computers” on AppleTV is unsearchable, and unbrowsable for large libraries
  6. TV Shows on “Computers” organize by show and season, not show (with season as a submenu), making it difficult to search. I have switched to Plex for TV Shows and Movies (but Plex has its own issues requiring renaming TV show files for compliance). I do use “Music” (i.e. Apple Music) for Music on the AppleTV, but don’t generally search my library with it and instead use Plex for my media. Searching by show and then season used to be better several years ago before an OS upgrade.
  7. Buffering issues on high resolution home movies streamed only on my home network, in the middle of the movie. Really? Even after switching to Plex. I can AirPlay a large HD movie file from QuickTime to AppleTV, but cannot stream the same file when requesting it via AppleTV (using either the Computers App via HomeSharing) or Plex. Why can I push but not pull?
  8. Two step verification is excessively requesting verification, asking me for verification on the same device multiple times, especially if I use a VPN. It seems to be related to cookie settings. There ought to be a better way so I don’t get the cookies I don’t want and do get the cookies I do. E.g. some way to identify verification cookies (from Apple, Google, DropBox) as distinct from ad-tracking cookies.
  9. iTunes has had much ink spent on it over the years, I will not add much more. I am sure they are working on it, and Apple Music seems to have settled down for me since all of my songs are now available, rather than just 25,000. Also I more or less gave up. But paying attention to Beats Music, rather than just getting the software working, is a misallocation of resources from the consumer perspective.



Problems that went away:

  1. Messages doesn’t work on one of my Macs at home. It works everywhere else. It just crashes on, or shortly after, startup. Why? [I will go with disk corruption on this … the hard drive subsequently failed. Fortunately backed up]
  2. Hangs on desktop. With an SSD Fusion drive. Why? [see above]
  3. Battery drain on iPhone 6s. I got rid of this once by wiping and reinstalling, but I don’t want to do that again. Some app is badly behaved I guess, but which? I have tried all the battery preservation tricks, and they seem to help, but I would like to have a fully featured phone (with background updates, and T-Mobile WiFi, and all my apps launched) without running out of battery in late afternoon. [I don’t know why, it just went away]
  4. Home Sharing hangs for some reason. This is on a local network through an Apple Airport Wireless device with a library on an iMac that is never asleep. I get messages like “Home Sharing is not available” and have to restart iTunes to get it back. [I stopped using Home Sharing directly as I switched to Plex for media].

The Best Show About Australian Politics That You are Not Watching

iurIn January, I highlighted the show Dreamland (Utopia), an “Australian show about the fictitious Nation Building Authority. So true to life, it is practically a documentary.” The star and creator of Utopia, Rob Sitch, had previously been associated with the team that made the movies The Castle (which is more important in Australian culture than The Big Lebowski is in America … “Tell him he’s Dreamin'”), and The Dish (Australia’s The Right Stuff or Apollo 13, sort of).

I just discovered a TV series Sitch co-created called The Hollowmen (presumably named for the T.S. Eliot poem). These are the advisors to the PM, never seen or heard, trying to drum up policy and running into all the constraints you would imagine. It is inspired by West Wing, Yes (Prime) Minister, and The Thick of It among others. Do be aware the Rob Sitch character is not quite the same character as in Dreamland, though you can imagine how they could have been sequels.

While I have less context to assert its specific truthfulness, it sure feels truthy, and is hilarious. So if you have completed Binge-watching Dreamland, go to the darker corners of the Internet (or Amazon – however not available without cracking your DVD player/DRM, and unavailable anyway) and get The Hollowmen.

Traffic Programming

Five-way intersection Franklin East River Road 27th - 05
Traffic Programmers will write algorithms control cars interactions with each others

Today we talk about traffic engineering, designing infrastructure, timing traffic signals, identifying where signs and markings go, and so on to improve the efficiency of moving cars (and ideally people) through the system (where the network itself is a given). This is distinct from transport planning, which identifies where infrastructure should go and what policies should be in place. This also differs from highway engineering, which looks at the geometric design of facilities, ensuring the horizontal and vertical curvature are safe (for human drivers at a given speed).  There are other related professions as well: urban designer, bridge engineer, pavement engineer, and so on.

The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.
The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.

One that we don’t have today, not really, is something we will need as a profession of the future: Traffic Programming. This will not be simply (or not simply) modeling human traffic, or re-arranging infrastructure, but instead will be design algorithms to control vehicles in real-time. This grows out of computer science, electrical engineering, industrial and systems engineering, and mechanical engineering, but will need to fully consider the environment around the vehicle and how to react to it. The early stages of this already exists, there are engineers and programmers designing autonomous vehicles. But they are doing so in a way that is autonomous — recognizing the existence of neighboring objects, people, and vehicles, but only optimizing for itself.

The Traffic Programmer of the future will design vehicles to do that but also communicate with neighboring vehicles, and coordinate with the system at a routing level as well. I firmly believe that effective Connected Vehicles will come after effective Autonomous Vehicles.

With appropriate price signals, we can do things now like aligning the system optimal interests of the network as a whole with the user optimal desires of the individual (autonomous or not) vehicle. Whether pricing can work to provide the right incentives on the give-and-take of traffic merging and lane changing is less clear. Drivers today negotiate that through eye contact and the actions behind the wheels. Finding protocols to do this automatically, and to negotiate directly between cars in real-time will be the task of the traffic programmer.