Posts about Autonomous Vehicles

Escalator reliability reaches 100%: Metro has achieved a new milestone for escalator maintenance. They have now reached a reliability rate of 100%; all escalators are currently broken at the same time. (Examiner, Matt Johnson)

Hop on I-395 PE: With Virginia's new program to sell naming rights to roads, Sudafed has proposed sponsoring all of Northern Virginia's congestion. (WBJ, Steve Offutt)

LOV-0 coming to a road near you: Google is reportedly working on a new program to design "passengerless cars," which will transport no people at all. In anticipation of this breakthrough, VDOT announced a plan to implement "Low-Occupancy Vehicle" lanes for their exclusive use. (Wired, Neil Flanagan)

DC4D4Thomas: DC for Democracy has endorsed Harry Thomas, Jr. as a write-in candidate for the Ward 5 special election. Members cited Thomas' consistency in talking about revitalizing the ward's main streets without making anything happen, creatively moving around money dedicated to serve youth, and his plan to solve transportation problems by setting up a series of Audi dealerships. (Geoff Hatchard)

Norton targets Wyoming: After several unsuccessful efforts to lobby state legislatures to support DC statehood, Eleanor Holmes Norton announced a new strategy to try to remove statehood from Wyoming, as it is smaller than DC. (DCist, Nick Clark)

If you want to cross a road, you'll just cross. The cars will slow down and avoid you. You could cross blindfolded and be perfectly safe. You'll be able to cross freeways. You'll be able to walk diagonally across intersections. You'll be able to do anything you want, and the cars will be responsible for avoiding you. Your biggest danger will come from cyclists and other pedestrians, not cars.

How much will they slow down? For how many pedestrians? Drum lives in Irvine, California, which has few pedestrians, so perhaps the cars can just avoid the occasional pedestrian. But in urban areas, there are a lot of pedestrians. If everyone crossed whenever they liked, the cars would slow down an awful lot.

In some places, cars would hardly ever get through. In almost any major city's downtown during a busy period, pedestrians are waiting in large numbers on street corners to cross. The only reason cars can get through is because signals govern pedestrian crossings. And when a light is green, often a car has to wait for a gap in the pedestrians or gently nose through to get past.

In Kevin Drum's future urban cores, constantly crossing pedestrians mean that car traffic will not flow at all except perhaps in the wee hours. Anyone who's been involved in a proposal to take away a lane of a road for bikes, or for a road diet, knows that drivers (or, in the future, car riders) will not stand for it.

Drivers are a powerful political force

Just look at, for example, the backlash against a bicycle lane on Prospect Park West in Brooklyn. In a very liberal jurisdiction, a modest and overwhelmingly successful bike lane nevertheless stirred up a few wealthy and well-connected individuals, including the wife of Senator Chuck Schumer (D-NY), to create an organization and file lawsuits to block the project using any means necessary.

Tea Partiers certain that there is a vast UN conspiracy to force them to live in high rises are opposing even extremely modest state laws creating some incentives for development in dense areas. Do we really think people will let government mandate that nobody is allowed to drive a car by hand, and that pedestrians get absolute priority?

In the DC area, some bicyclists ride on MacArthur Boulevard in Potomac, a narrow and windy road in a low-density area. That's perfectly legal, but there's a constant stream of letters to local press outlets by drivers who are sure it must be illegal to bike there since it slows them down.

Forcing drivers to travel slower would be like telling seniors that we're cutting their Medicare. The political counter-pressure is intense, so much that most transportation planners always take great pains to reassure drivers of how any change won't really slow them down. Even for the pedestrian plaza in Times Square, one of the early promises from the mayor's office was that it would actually reduce car delays.

I can go on. But anyone who writes regularly about transportation has encountered the massive sense of entitlement from drivers. When I'm driving, I hate to be delayed, too, but I squelch this natural impulse because I write about the issues and have context.

It may well come to pass that driverless cars have to travel slower and pedestrians are able to act more freely. But this will create tremendous political pressure to change the social compact over roads to get traffic moving faster once again. And in this, we will see another, more intense variant of the same fight we have today.

Once, pedestrians did walk freely, and children played in the streets. As automobile use proliferated, rising deaths led to campaigns to segregate street space. Our society could have taken one of two approaches: it could have limited drivers, and added legal liability to force drivers to be more careful, or it could get people out of the street. Many places in Europe chose some elements of the former, but America decisively chose the latter: to redefine the street's role in society to move cars faster. I'm certain that in Drum's scenario, there would be intense pressure to do the same.

Who is liable?

One element determining whether driverless cars turn into the Kevin Drum reality or another one is how we treat liability. When a driverless car kills a person, whether due to a human overriding the technology or a failure in the computer system, there will be a lawsuit.

If courts hold that the manufacturer of the car is liable, this will stifle development of the cars. The technology might ultimately be perfect, but it won't be perfect from the start. Manufacturers will ask state legislatures to limit their liability. Already, a number of commentators have called for liability caps or other legal changes which shift the burden away from the manufacturer.

If the legislatures don't agree, then manufacturers will have to move very carefully until they can make the cars virtually incapable of killing anyone. That will likely hinder development in general, and make any self-driving cars travel slower than human-operated cars. Many drivers therefore will turn off computer mode a fair amount of the time, and political pressure will build to change the liability standard. This will be an early skirmish in the battle over the cars' speed.

If states do limit liability, then we'll end up with a different situation. Buyers will want driverless cars that use algorithms like the one the University of Texas team devised that let them move faster. Sometimes those cars will travel close to pedestrians or bicyclists. Most of the time they'll still avoid killing anyone, but mishaps will happen. And like in today's legal world, prosecutors, judges and juries will be very reluctant to impose heavy punishments on someone operating a car who unintentionally kills another.

Then we'll be back to a situation like the early 1900s roads. For people's own safety, officials will start imposing restrictions on pedestrians. It'll start in places like Irvine. If laws won't stop people from walking on highways or crossing diagonally, then they'll build fences, or skybridges, or both.

Today, one argument against restricting pedestrians too much is that not everyone can drive. Seniors and people with disabilities can't operate a car, and many can't afford them. When driverless cars become commonplace, there will also be cheap taxi service, and so it'll be easier just to tell people to call up a car.

Already, many suburban areas are essentially an archipelago of human-accessible islands in a sea of almost-cars-only space. Little will stand in the way of making this other space absolutely cars-only. And why not? After all, without people, cars can use fancy algorithms to interweave with each other and zoom around far faster than they could in 2012.

Driverless cars aren't bad

A number of the responses seem to be reacting to an imaginary variant of my thesis, in which I said that self-driving cars were going to be a unmitigated bad thing. There's a natural tendency to simplify all arguments into "x is great!" or "x is terrible!"

The fact is that autonomous cars are coming whether we like it or not, and like any technological advance, will bring both terrific improvements to people's lives as well as drawbacks.

Driverless cars are sure to lead to big fights. Will they shift the balance farther toward pedestrians, as Kevin Drum believes, or away? I hope the former, but the technology won't magically solve this problem. Instead, we'll have to fight it out through the democratic process, as we do most other issues affecting the public sphere.

Badger writes,

[H]uman-driven cars would have to wait for a signal that would be optimized based on what everyone else is doing. And the same would be true of pedestrians and bike riders.

My background is in computer science, too, and computer scientists love figuring out how to make complex systems perform efficiently. Driverless cars provide an opportunity to optimize the real-world traffic system, if you can get most people driving computer-controlled cars and can get all of those computers to cooperate.

But you can't optimize people so easily. Already, cities host ongoing and raucous debates over the role of cars versus people on their streets. For over 50 years, traffic engineers with the same dreams about optimizing whizzing cars have designed and redesigned intersections to move more and more vehicles.

These changes frequently pushed other users aside with longer waits for crosswalks, the need to push buttons to get a walk signal, awkward bridges over wider and wider arterials, or simply omitting bike or pedestrian facilities entirely and then blaming those users when careless drivers hit and kill them.

Some pro-automotive advocacy groups like to push the theme of a "war on cars," but bicyclists and pedestrians feel like there's been a war against them since the early 20th century. This Texas team's video just perpetuates that impression.

The video even depicts an intersection with a whopping 12 lanes for each roadway, at a time when most transportation professionals have come to believe that grids of smaller roads, not mega-arterials, are the best approach to mobility in metropolitan areas.

Driverless cars, therefore, are poised to trigger a whole new round of pressure to further redesign intersections for the throughput of vehicles above all else. It won't only happen in the cities, either. Suburban areas are often ground zero for these debates, where the majority of people drive, but a significant and often growing number are either unable to drive due to age or disabilities, or are unable to afford cars. (Driverless cars probably won't be cheaper.)

Suburbs, therefore, often develop a greater tyranny of the majority, where county and state departments of transportation optimize their roadways for car throughput and leave bus stops in awkward and narrow roadside spots, leave crosswalks out or even remove existing ones, and set the stage for rising deaths.

If autonomous cars travel much faster than today's cars and operate closer to other vehicles and obstacles, as we see in the Texas team's simulation, then they may well kill more pedestrians. Or, perhaps the computers controlling them will respond so quickly that they can avoid hitting any pedestrian, even one who steps out in front of a car.

In that case, we might see a small number of people taking advantage of that to cross through traffic, knowing the cars can't kill him. That will slow the cars down, and their drivers will start lobbying for even greater restrictions on pedestrians, like fences preventing midblock crossings.

Our metropolitan areas could then look, more and more, like zoos for humans interlaced with pathways for the dominant species, the robot car. Maybe the machines really are on the way to taking over, but instead of Skynet declaring war on humans, we'll be the ones passing laws and reshaping our communities for their convenience.

I'm not suggesting we avoid research into driverless cars. Like any technology, they can bring good or evil, depending how society handles them. Driverless cars can allow buses to become on-demand jitneys and virtually eliminate the need to own a personal car in a city, or to build huge amounts of parking under office buildings. Instead of storing cars during the day, they can just drive around and transport people like taxis.

But we do need researchers excited about driverless cars not to forget the human element. The goal of our built environment is not to move cars as fast as possible everywhere, but to create a better quality of life. The computer science researchers need to also talk to their colleagues in other disciplines, set appropriate goals that consider all users of the roads, and think about what algorithms can actually make life better.

Cross-posted at The Atlantic Cities.

If it's not needed, a self-driving car can park itself at an offsite location, thus eliminating the need to build large amounts of parking at desirable (and expensive) locations.

They'll reduce labor costs: A self-driving car needs no operator, thus removing human labor from the equation. Self-driving cars will put taxicab drivers out of business. What will those thousands of people do with their skillset when a computerized system makes them obsolete?

They'll expand access to transportation: The process of driver training and licensing will be obsolete, and the requirement that people be 16 or 18 to drive a car will be irrelevant since now there are no drivers, only passengers.

This is great news for the disabled, especially the sight-impaired, as well as for adults who have lost the ability to drive. Will we create some new paradigm of age restrictions for being an unattended passenger?

Self-driving cars eradicate the car-ownership paradigm. If you can easily and affordably (remember, no labor to pay) book taxi service from your smartphone, more people than ever will eschew the costs and annoyance of car ownership.

They'll be safer: Self-driving cars likely won't make human errors. Auto crashes typically claim around 38,000 lives per year, and that's been true for decades. Over 80% of these are attributable to human error, either negligence, distraction, incapacitation, malice or other uniquely human quality.

They'll reduce congestion: Self-driving cars can manage congestion as a system, rather than a collection of self-interested units. A lot of congestion stems from the way each driver acts in his own self-interest. For example, changing lanes might (or might not) help one individual driver, but hurts the overall performance of the road. Speeding into a gap and then braking also creates worse congestion overall.

If all cars are self-driving, then they can cooperate to mitigate congestion. For instance, the cars could all slow down to 35 mph past a crash or police traffic stop, rather than allowing the speeding up and slowing down and rubbernecking which lead to traffic and more crashes. Over time game theory and other disciplines will help engineers devise ever more complex strategies to keep the system performing optimally.

They'll make current transit economics obsolete: Self-driving cars represent a major existential threat for current and planned transit systems. Our current transit paradigm relies on capital and operational subsidies. We can't charge riders enough to pay for everything that goes into making transit work. As we raise fares, more riders forego transit and choose the automobile.

If, as I suspect, self-driving cars are handled primarily in the private sector, their operations will not be subsidized, and their relative convenience and utility will call into question the logic of investing billions into the construction and operation of transit systems.

They won't last as long: Automobile manufacturers will have to adapt the volume of vehicles they produce annually. While many fewer cars will be needed across the economy, those that are autonomous will be driving much more frequently. Their replacement cycle would be more similar to police vehicles, which only last around 3-5 years before wear and tear makes replacement a better option than repair.

Most passenger cars today spend around 98% of their time parked somewhere in between single-occupancy trips. Consequently, their average lifespan is between 15 and 26 years.

They can be electric: An electric self-driving car can go to where the charging stations are. DC and other governments are currently embarking on a campaign of spreading electric vehicle charging stations around the urban environment under the assumption that we must cast a wide net of these kiosks around so that they are convenient to an EV owner's origin or destination. But in a few years, it is likely that this will be entirely unnecessary, and rather the car can take itself to a central charging location, like a power substation or electrified parking garage that can efficiently charge hundreds of vehicles on an as-needed basis.

They'll change culture: A self-driving car eradicates a unique part of the American identity, the freewheeling mastery of the open road. We'll wax nostalgic for what we've lost, but everyone will benefit from the gains.

A world with self-driving cars would operate very differently than the one we currently live in. I would say that's mostly for the better. As urbanists, we've often succumbed to a gut reaction that cars are bad, transit is good. However, the reality is that it is not cars that are bad, but the single-occupancy driver paradigm that is so damaging to our environment, urban fabric and quality of life.

We still live in an America where 78% of people drive to their jobs by themselves. I'm convinced that we're about to see that start to change as self-driving cars become a reality. It is time to start having the conversation about how we want this future to unfold in order to best plan for a very different world.

In only 32 years at the turn of the twentieth century, the primary power source for human mobility for 7 millennia became obsolete. If there was a comparable transportation planning field at the time, using our current system of forecasting, their models from 1890 would have estimated a linear increase in horse trips for the next 30 years, with some additional subway and horse-drawn streetcar lines.

They would not have imagined that there would be no horses at all. Instead, they would need a complex set of signal and traffic control technologies to address the new safety and mobility issues presented by automobiles.

Today we are at a similar inflection point. The occasional news story in Popular Science or the New York Times describes the wondrous technology that allows cars to operate under complex scenarios without any driver input. However, we don't hear a peep from transportation planning organizations about how society will adapt and plan for this change, even though it seems increasingly imminent.

Washingtonian recently interviewed Michael Pack, the region's foremost traffic technologist. The interview is alarming: "I ask how long before we can all stop driving and let the cars do the work," the interviewer asks. Pack responds, "Oh, a while, Maybe 30 years." Pack appears to be considering a network of interconnected cars that can talk to each other and have a situational awareness that allows them to travel at 65 mph and 6 inches from one bumper to the next.

If Pack's estimate is reliable, then that means we should start seeing the first generation of autonomous cars that replace the human driver with superior situational awareness in a matter of years. It also means we are already within the 30-year window where transportation planners should anticipate adoption of this technology and its consequences.

My best guess, based on publicly available information, is that within 7-12 years, there will be a commercially available autonomous vehicle sold in the US

In sharing this theory, I've heard my share of skepticism. But most people would accept on faith that within 100 years we'll have autonomous vehicles. Some would accept that they'll be here in 50 years, while few would accept that we'll have them in 5 years. So at heart, the discussion is not a question of if this technology will develop, but when, and whether we have to start thinking about it from a policy and planning standpoint.

Whichever manufacturer is first to roll out a consumer-ready version of this technology will have a blockbuster product, so the economic incentive to be first is enormous.

In a future post, I'll discuss some of the changes we're likely to see in a world with self-driving cars.

Today, buses have fairly high operating costs because of the labor involved. They're far more space-efficient than cars, but it's expensive to pay a driver. The biggest reason transit costs more to operate than highways is because with transit, you're paying a driver, but if you're driving your own car, you're doing the labor yourself.

That's a structural imbalance that puts transit at a budgetary disadvantage, until your car drives yourself and so can the bus. Suddenly, there's no difference.

Cities with self-driving trains, like Vancouver, can run far greater frequency at low ridership times like nights for a reasonable price. More vehicles makes transit more appealing.

Now, buses are often very infrequent or nonexistent at night, making them an unappealing mode for an evening trip, for example, unless you live right near a high-frequency, late night line. But what if the bus just drove itself every 5 minutes? Basically, it's like everything PRT promises, but without any guideways.

Or, better yet, what if it ran on demand? Then it's like a cheap taxi. And, in fact, buses, taxis, and car sharing will essentially merge into one mode.

Right now, each of those modes has advantages and disadvantages. Buses have the labor cost issue. Taxis have it even more, but are very convenient when available. Car sharing is great for certain types of trips, but you have to return the car to its starting point.

With autonomous vehicles, there will be no need to differentiate between vehicles that carry many people on fixed routes (buses), vehicles that carry few people on demand (taxis), and vehicles you can drive but are only in certain places (car sharing).

Instead, you'll simply be able to call a number or use a mobile app to book a trip. In urban areas, you'd go to a designated bus stop, or maybe pay more to get a custom pickup right where you are. A vehicle will show up at an assigned time, maybe picking up a few other people as well who are going to a similar destination, unless you want to buy a solo trip.

For more on how this could work, see Mark Gorton's Smart Para-Transit articles.

We'll need far fewer parking spaces in dense areas. Commuters from suburbs that take autonomous buses/taxis/paratransit vehicles/car sharing cars into the city won't need to park them. Instead, they can drive themselves around all day serving short-range trips as taxis. Others would live in the city full-time, and those might need overnight parking, but far less than we have today.

It will simply become far easier and cheaper to live in the city without a car. It won't really be like living without a car today. It will simply be like living with a car, without worrying about parking or paying nearly as much. Or, perhaps, it's like using car sharing, but with the guarantee that a car is always available and without having to worry about getting the car back in time.

Finally, assuming Google can perfect the software, streets will become much safer. The cars will be far better at avoiding crashes. Instead of most drivers speeding, turning improperly across bike lanes, and being distracted with texting or phone calls or the radio, the car will be "paying attention" at every moment (as long as its software isn't buggy!)

Governments and manufacturers will have to decide how to handle speed limits. Today, many localities set speed limits knowing most people speed. Will drivers be able to choose their speed, or will cars be required to travel the actual speed limit? If so, maybe we can finally rationalize all these limits and set them appropriately.

If governments set up autonomous car-only lanes on freeways or ban human-driven cars altogether once most cars drive themselves, traffic can move more efficiently. Right now, if a bunch of cars are stopped in traffic or at a light, each driver has to wait a few seconds after the car in front starts moving. With computers, the entire line of cars could simply start moving all at once.

The greater efficiency will allow existing freeways to carry more people than they do now, and since it'll be a lot easier to create carpools with Smart Para-Transit dispatching, even more if jurisdictions are willing to convert lanes to HOV. That could also accelerate sprawl, but the greater ease of urban living and lower need for parking will also facilitate infill development as well.

How else will cities change if cars can drive themselves?

This reminds me of the Google shuttle: Around 2003, Google set up regular Wi-Fi-equipped shuttles from San Francisco to their Mountain View headquarters. Suddenly living in the city became more convenient for the many employees who desired the city life. At the same time, the shuttle also made it more feasible for Google to remain in the office-park wastes without losing urban employees to San Francisco-based startups. Might some employers move out to the suburbs now that their city employees can take a robocar to work?

To take this to an extreme, if we all had personal teleporters, some people might choose to live in the city, but from a pure cost standpoint, few employers would if we continued to underprotect undeveloped wilderness. On the other hand, unlike with teleporters, the road network's capacity constraints wouldn't go away. At most, this might be a gradual shift and dependent on our governments' appetite for more roads.

Autonomous cars would change transit as well. Rail wouldn't change much (until we have robo-tunnelers), but buses would become enormously cheaper to operate without drivers. And without drivers, why hew to fixed schedules? A fleet of buses could dynamically pick up people as they enter their locations and destinations on mobile phones—the Smart Para-Transit concept Streetsblog's Mark Gorton wrote about.

With a comprehensive, and cheap, Smart Para-Transit system, perhaps the best hope for more transit capacity would shift to roads, and transit advocates and suburbanists alike would start advocating for roads together. Of course, with the safety and efficiency of robocars, we wouldn't need freeways, and avenues like Connecticut could move even more people while killing fewer.

Ultimately, robocars won't stop us from having suburbs and cities. But they'll give people more choices. For those who want a big house, it'll be a bit more feasible (though still energy-costly); for those who want to live in a walkable neighborhood, it too will become more realistic thanks to a reduced reliance on capital-intensive rail construction (assuming zoning doesn't prohibit new dense development).

And choice is key. When I started this blog, having recently read The Option of Urbanism, I wrote, "Too many people are forced to leave great neighborhoods to find affordable housing or good schools. If people want to live in single-family homes, they certainly may. But everyone should have the choice to live in an apartment or townhouse in a walkable, safe, livable neighborhood. ... We must ensure that there are enough great neighborhoods for everyone who wants to live, work, shop or play in one."

If people want to drive, that's fine, as long as they're not foisting big surface parking lots, danger, and pollution on the city dwellers without paying the cost. If robo-cars (and the more sci-fi orbital solar stations providing unlimited free energy) eliminate the negative externalities of driving, then we really can all have a better and freer choice.