in , , , ,

Forget Smart Cities, ‘Stupid’ Infrastructure Is The Solution For Future Transportation

As the potential for a digital revolution in transportation has risen, many people have pondered the question of how we might make our roads and cities “smart” to enable this revolution, and what types of new infrastructure we should build to be ready for it.

The answer I give, which naturally disturbs those in the infrastructure world is, “not much.” The true answer is not so facile, but relies on several fundamental lessons from the core digital technologies which have already vastly changed the world, computers and networking. That’s the world I come from and it has worthwhile lessons for all those trying to follow on the exponential revolution pathway.

The core lesson is to keep your infrastructure simple and general, rather than imagine you understand the “application” that will use it enough to try to solve its problems. Infrastructure changes at the pace of decades, while digital technology changes daily. You can’t plan for 2030s digital cars with the knowledge of 2021 — you can try but you will almost certainly be wrong — the best you can do is make things flexible and delay decisions to make them with the knowledge of the 2030s. That’s quite a change for city planners used to writing plans for 10, 20, 30 and even 40 years out.

In 1997, David Isenberg published the important essay, “The Rise of the Stupid Network,” in which he details how the internet took over the world by being as simple as possible. Name derives from the 1980s marketing name the telephone companies had for their new systems, “Intelligent Network” or “IN.” The IN tried to put the intelligence for new phone functionality in the network, in the infrastructure. The phone company designed, built and managed innovation in telecoms. You connected with your standard plain old phone. (Younger readers may not know it, but everybody back then had a basic phone on their desks with wires coming out of it which you used to talk to other people.)

When the smarts were in the infrastructure, we relied on the infrastructure for the innovation. And why not, the Bell Labs scientists were among the best in the world?

The internet flipped that upside-down. The core design of the network is dead-simple. In fact, it’s essentially the same design today as 40 years ago! Even so, we’ve seen the greatest period of innovation in human history on top of that stupid infrastructure, and it’s not a coincidence. On the internet, all the smarts are in the edge devices. Your phone. Your laptop. The web server that sent you this web page. Everything is there, even the negotiation of network link quality and speed which you might imagine should be in the infrastructure, which is much closer to those factors. The internet itself just delivers postcards from A to B, really fast. Its only job is to figure how to move those postcards.

When the smarts moved to the edges, they got a lot smarter. Anybody could innovate. Nobody needed the phone company’s permission, the way it used to be. A few folks in Europe wrote a program called Skype which took over most the world’s long distance business for a while. They didn’t ask the network companies to get involved or even give permission to eat their lunch, and they certainly would not have received it.

As noted above, the problem is that today, you can’t know the future. People in the computer industry have gotten used to that idea, where the capabilities of the computers and networks have been doubling in performance every 1.5 years for over 5 decades. You can’t plan for 2030 in 2021 so you don’t. Instead, you keep what you must build simple and put as much as possible into software. That’s because you can change all your software in 2030 when you learn the reality of the future, and it’s free to deploy it, even though not to write it.

Waze and the Cell Phone

If you think this lesson is only for computer networks, consider that a decade ago, every transportation planner was keen to measure traffic. Taking the smart infrastructure approach, large sums were spent installing magnetic coils and traffic cameras to report on just a few streets. Then a small team in Tel Aviv wrote and gave away Waze, and suddenly had great traffic data (and more) on every road in the world, for free! They let the road remain stupid and relied instead on the wonder device of the 21st century, the mobile phone.

The mobile phone is amazing. Your typical smartphone today is around a year old. It was designed less than a year before it came out. The software on it was written last week. You’ll get another one soon and just stick the old perfectly working one in a drawer. Compare that to cars which last 20 years and are designed 3-7 years before they come out, and until recently, software that was years old as well. Or to roads and infrastructure, which must last for many decades and broadly follow designs of centuries past.

If you have an incredibly brilliant plan for a smart infrastructure improvement, you will lose to the cell phone. You won’t imagine you will because today’s cell phone doesn’t even make any sense for your problem. But in 10 years, when you are getting some deployment, we will have gone through 10 generations of cell phones. They will perform hundreds of times better with new sensors nobody is aware of today. A year after that they will be even better.

Unless you have a proof from physics that the digital devices can’t attack your problem better, they are going to win. Hugely. Immensely. It won’t even be funny.

Internet people are used to this approach. The people who design internet routers give no thought to supporting Facebook. Even the people who design your web browser, the most important platform for Facebook, give few thoughts to the specific needs of that site. We encourage them not to, and call it a “layer violation” because we believe strongly in the idea of independent layers, and that philosophy has served us well. People who build internet infrastructure would feel not affront at being told they should not be working on making their “roads” smarter for Facebook to use. If you suggested it to them they would think you were nuts.

Virtual Infrastructure

High-tech transportation companies have been moving to the idea of “virtual infrastructure,” which is to say a sort of digital twin for the physical. The most common version is digital maps, which are used by almost all self-driving car teams to improve performance and safety. But virtual infrastructure also includes data on traffic, road hazards and even traffic signal timings. In some future world, which is not coming soon, where human drivers no longer drive the road (or they all drive with a heads-up augmented reality display) there need not even be lane lines, road signs or traffic signals — these could all be moved to the virtual.

As noted, that’s not happening soon, but consider the power if it did — the ability to restripe the roads, change the direction of traffic, reallocate road space and parking with the push of a button. Roads might have most of their lanes going downtown in the morning, and some of them one way going out in the afternoon if traffic rules all existed in the virtual layer.

Parking lots certainly could benefit faster. Self-parking cars don’t need to use regularly marked parking spaces. They could be given a section of the parking lot only for them where they park in a denser-than-valet style which fits 2-3 times as many cars in the same space, all just with software.

The ideal road, it turns out might be just bare pavement. Its only inherent characteristics would be its dimensions and the weight it could bear — more on that below.

The first law of robocars

The use of maps led me to declare in 2007 what I called “the first law of robocars,” namely “You don’t change to world to fix your problems.” You must drive on the road you are given. There was zero doubt about this law in the beginning. The idea that small, scrappy teams of programmers could ever ask the world to build special roads for them was ridiculous. Later, as the world caught the fever of self-driving cars, and government transportation agencies and major automakers all fully embraced it, they got seduced by the idea that things were big enough now that you could consider it.

The less advanced teams, building things like ADAS lanekeeping or simpler cars, tried to track painted lane markers and found they were badly maintained. The call went out, “You need to paint the lanes better to help the self-driving cars!” But that call didn’t come out from the leading team at Google, where I was working. Our car built its own map of the lanes and it actually preferred broken and messy lane-markers because they helped it recognize better just where it was on the map. While it was able to handle a freshly painted road, it gained extra confidence knowing it was in exactly the place where the left marker is broken 2/3rds of the way down, and there is a repair crack shaped like a duck on the right side. It didn’t want or expect the road to be repainted for it, but instead adapted to the road it was given.

But we need more capacity

Many cities are plagued by congestion, suggesting that we must build more and smarter infrastructure to relieve this. While transportation planners often feel that new infrastructure fails to do this due to induced demand (the new capacity is quickly filled up by new users attracted to its extra capacity) they often hope to deal with the problem with specialty infrastructure, such as dedicated transit right-of-way.

At the same time, when we look at the roads, even the most congested ones, we see that most of the seats going by are unoccupied. There is clearly 2-3 times as much capacity just in the roads and cars out there now. Indeed, if the roads were occupied by full vans at rush hour, their capacity would be astonishing, far greater than any other transportation system including the best heavy rail lines. Streams of bus-sized vehicles on a single lane could move 150,000 seated people per hour if you needed to, if given the right software and offline stations.

If you consider a world where nothing is built in, where the directions of lanes and streets can be flipped in software, where what’s parking and what’s road can change with the hour if desired, you will discover there is far more capacity in our existing infrastructure than previously imagined.

If we do build new infrastructure we can make it simple and cheap. If desired, houses in low traffic areas can have roads barely wide-enough for a fire truck, with wider spots for passing every so often, made use of by cars that time their passing perfectly to always meet in them. New infrastructure can be made only for small and light vehicles, leaving the dwindling number of heavy vehicles to use the older roads. (Without drivers, you don’t need 40-ton trucks when 8 5-ton trucks can do the job in the city 99% of the time, doing far less damage to the road and needing much less of it. We have yet to even dream of some of the things we’ll do in infrastructure whose structure is mostly virtual, and dynamic.

Some Lessons

Lessons of the value of simple infrastructure are everywhere in the internet world. However, here are some worthwhile examples out of transportation infrastructure.

Online stations vs. Offline

A virtue of stupid infrastructure is you don’t define functionality like lanes and stations in advance. Contrast that with typical rail lines, which make two limiting decisions. The first is to use rails. While pavement carries pedestrians, scooters, cars, bicycles, trucks, vans, buses, robots and vehicles yet to be invented, rails are designed to handle trains, trains, trains or trains. In addition, because it’s harder for trains to change paths at will and go around other trains, we often see rail built with “online” stations, where the train in the station blocks the line. This forces long headways between vehicles, measured in minutes, compared to the 2 seconds used by human driven buses and the 1 second possible with robots. It also makes it much harder to have express vehicles.

This is one of the many factors of simple infrastructure that answers a frequent question about LRT.

Light rail vs. BRT

Many cities seek to install Light Capacity Rail, but may contrast it with Bus Rapid Transit. LRT is sexier, usually electric and often given private right of way and larger cars which makes people think they want it. But it requires infrastructure dedicated to it, and is almost always done with online stations. BRT, on the other hand, uses simple infrastructure. Aside from being vastly cheaper than LRT for the same capacity, it is much more flexible. You can change your mind with BRT and reuse the pavement for other purposes with no or minor changes when the future.

In a way the ability of BRT vehicles to go out onto pubic streets is a great strength, but the temptation that provides actually scares people away from it. Once BRT enters the public street, it is like a bus and subject to traffic, while dedicated LRT right-of-way will pretty much always be like that (at much higher cost.) But the ability to adapt to the future is worth it. Indeed, dedicated LRT lanes might make sense to allocate to future automated transit vehicles to fill up the capacity of the lane. One can also make managed lanes on public streets where some private vehicles are allowed in the managed lane — enough to do a meaningful increase of capacity but not so many as to slow down the BRT. This can be done particularly well with future cars either with automation or techniques to assure no driver dares to let a bus get so close as to have to brake (causing a phone and expensive ticket.)

Elevators

Elevators are dedicated infrastructure, but in recent years, people have been able to turn them into software controlled devices to great benefit. In many cities, most tall buildings don’t have buttons in the elevators any more. Instead, people select their floor before getting on, and are told which elevator is going to that floor. This groups together all the people going to the same floor, making the trips have fewer stops and significantly increasing the throughput of the elevator with just some software. In the future, algorithms may get better, and notice people approaching the building to plan ahead. (Some buildings summon an elevator for you to your floor when you scan your badge at the entrance.)

These new elevators have allowed buildings to get by devoting less floor space to elevators for the same volume of traffic — using software tricks actually has increased the capacity and expanded the usable area of new buildings, or allowed them to be narrower.

The approach of these elevators offers a lesson for transit, especially if, unlike elevators you can have offline stations. (Offline stations are a dream of elevator designers which would massively increase elevator capacity and thus building floor space.) Imagine a transit line where instead of 10 cars in a train, there are 20 vans waiting at stops, and you say what station you want (on any line) and are told which van to get into. On that van are a bunch of people going to that station, or sub-groups going to a small subset of stations, and off you go for a non-stop or nearly non-stop trip, even when changing lines.

DSRC vs. C-V2X vs ?

These acronyms come from the wished-for world of communications from cars to other cars and infrastructure, known as V2X. V2X planners hoped for their own dedicated spectrum and radio hardware — special purpose infrastructure. Their first effort, DSRC, was designed in the 1990s and is now losing its battle (even before it got deployed) to C-V2X, which is still a special-purpose system that just uses newer technology and piggy-backs on cell phone tech. The reason for that loss is clear — who is going to use a radio system designed 25 years ago? Do you use your 25 year old cell phone very much? No dedicated system, no matter how nice it might seem, will compete with the march of the public data networks, which get a new generation every few years.

Above, I described the internet’s philosophy of “layers” where any attempt for one layer (like applications) to try to dictate or interfere with lower layers is called a “violation” for good reasons. You let the data communications companies worry about making good data communications infrastructure, and to constantly be improving it. You expect it to get a new generation every few years, and to be available in phones long before it shows in cars. We can now guarantee that they will do this, and for their own reasons they will provide all the capabilities sought — reliability, low latency, high capacity and many more nobody even knows to ask for today. You build your own layers on top of that, doing everything in software so you can change it on a regular basis. You try not to standardize, allowing instead for constant change an innovation.

Keep it simple, stupid

Nobody has a crystal ball to give them the right plan for the 2030s. If you have a plan for the 2030s only one thing is certain about it — it’s wrong. The world is going to get more and more like the world of computers as time passes, and every field will need to learn the lessons the worlds of computers, networks and electronics had to learn the hard way. Keep it stupid.

What do you think?

New Advances In Regenerative Medicine

The Effects Of AI On Child Psychology