Impressive advances in vehicle automation technology, together with enthusiastic media coverage, have sparked eager anticipation of an imminent revolution in urban transportation. Driverless taxis (or automated cars in car-sharing fleets) are widely expected to play a central role in this revolution.
Imagine: automated systems driving far more safely and efficiently than humans—and eliminating the considerable expense of taxi driver wages. You travel door-to-door in privacy and comfort, and at low cost. Public transit as we know it becomes a relic; private vehicle ownership dwindles away. Car designs mutate and urban land uses are reconfigured. This revolution is said to be just a few years down the road.
On closer examination, driverless taxis would indeed enhance urban mobility, but because of technological limitations, they’ll be most useful where population densities are lower. Elsewhere, buses and trains will still be essential. And driverless taxis are likely still some distance away—some experts caution it may be decades before cars are capable of driving themselves in the full range of complex and chaotic urban environments. But this is not a negative prognosis for vehicle automation. Before driverless taxis emerge, less advanced forms of automation—some are already here, others are coming soon—could improve safety and ease travel for motorists in private cars. Small, lightweight shuttles that operate fully automated at low speeds could fill a useful niche in low-risk environments. And even the less advanced automation technologies could significantly improve bus service.
Driver wages make up the majority of taxi operating costs, so the cost of a journey would plummet if a driver were replaced with an automated system, provided that the cost of the system is low enough. Meanwhile, a similarly equipped vehicle in a car-sharing fleet could chauffeur a traveler to their destination and then navigate to the next traveler, thus providing service identical to driverless taxis. On-demand, private, door-to-door, inexpensive travel—one estimate is that a fare for a five-kilometre driverless taxi trip would be similar to bus fare—would also come without the hassles of parking, insurance, maintenance, and the expense of purchasing a vehicle. Many travelers would be enticed away from public transit use and from private vehicle use and ownership. Broader impacts would follow: for example, parking demand would shrink as cars serve multiple travelers rather than sit idle, thus freeing up valuable space and stimulating a comprehensive rethink of urban land use.
How large of a role could driverless taxis (or automated vehicles in car-sharing fleets) play in a future urban transportation system? A recent study explored a scenario where all trips in Singapore would be served by automated car-sharing vehicles. The researchers found that travelers would wait, on average, less than 15 minutes if one car was provided for about every four households. That’s around one car for every 16 people. It’s a remarkable figure, but a questionable one, because it rests on the assumption that the automated cars would move at the same speed as traffic today.
Most Singaporeans use transit, so a wholesale shift to automated car-sharing vehicles would cause a huge increase in traffic—and a major drop in speed. But wouldn’t automated cars share the roads far more efficiently than human drivers, yielding faster and smoother flows? It’s true that with their precise control abilities and short reaction times, automated cars could drive in close formation and increase throughput, but this potential comes with conditions and limitations.
For one, automation needs to be teamed with vehicle-to-vehicle communication (V2V) to greatly improve flow. To safely drive at short headways, an automated system needs to know the movements, braking capabilities, and other properties of several cars ahead. With V2V, vehicles can share that data wirelessly with each other. In contrast, automated cars that rely solely on on-board sensors are blind to much of that information, and they detect movements of the vehicle ahead with longer time lags. Consequently, they can’t drive safely or smoothly at short headways.
Automation with V2V, on the other hand, can significantly shrink the gaps between vehicles. Some studies have found that road capacities would quadruple or even quintuple; however, these studies modeled uninterrupted streams of cars. In a more realistic scenario, cars follow each other in tight groups, with the groups, or "platoons," separated by larger gaps. Inter-platoon gaps improve safety and allow vehicles to enter and exit the freeway. It’s estimated that platooning could increase a lane’s capacity by 50 to 100 percent.
On city streets, however, platooning would be infeasible. Flows there are much more complex: cars are making left turns, parallel parking, and stopping for pedestrians and red lights, to mention just a few manoeuvres. In a different proposal for decongesting streets, automated vehicles would communicate with a roadside computer that safely orchestrates their intricate web of movements through an intersection. No stoplights or stop signs would be needed. According to simulations, such a "reservation-based" intersection could accommodate almost as much traffic as an overpass. This kind of performance is unlikely, however, when pedestrians, cyclists, and non-automated, non-communicating vehicles are added to the mix.
Similarly, before platooning-capable vehicles dominate the freeways, capacity increases will be small. Studies suggest that when less than a third of vehicles on the road are outfitted for platooning, capacity would increase little—large gains come only after two-thirds are equipped. This means that even if adoption is rapid after platooning-capable vehicles appear on the market, sizeable capacity improvements could lag by two to three decades.
Driverless taxis could also generate new traffic. Empty taxis driving between passengers could add 20 percent to vehicle miles traveled. And the ease of automated travel would encourage people and businesses to locate away from urban centers, thus leading to longer trips. More hopefully, transportation expert David Levinson suggests that driverless taxi users would be incentivized to choose more central locations to reduce their fares.
If driverless taxis were smaller than current cars, though, that could improve road capacities. Suppose that automated cars have excellent crash avoidance capabilities: heavy vehicle designs would no longer be necessary to protect occupants from impacts. Cars could be smaller and lighter, and therefore occupy less road space, cost less to manufacture, and use less energy. Compact cars serving short taxi trips would also facilitate a shift to electric drive. There would be no need for steering wheels and other manual controls; thus, interiors could be fully redesigned, making driverless taxis even more attractive. The catch is that small, light vehicles would be vulnerable as long as heavy vehicles are around—especially heavy vehicles driven by humans. Until crash avoidance is sufficiently advanced, or until heavy non-automated vehicles are rare (or traffic is segregated), a shift away from crashworthiness is unlikely.
Even when platooning begins to produce major capacity improvements on freeways, buses will be indispensable in dense urban areas. This is simply because buses occupy far less road space per passenger than cars and can thus move more people through narrow corridors. Where densities are lower, though, the space efficiency of buses is less critical. In these areas, driverless taxis could feed riders to higher-capacity bus and train routes and provide inexpensive, on-demand, door-to-door suburban trips. Individuals could travel in privacy, without owning a vehicle—or multiple travelers with compatible pick-up and drop-off points could share a cab, a possibility that’s been examined especially by researchers at Princeton.
When might driverless taxis hit the roads? Some pundits contend that the necessary technology, dubbed Level 4 automation by the U.S. National Highway Traffic Safety Administration, will appear very soon. For example, it’s widely reported that Nissan is promising Level 4 by 2020. However, these reports misinterpret Nissan’s announcement about their "Autonomous Drive" feature, which will not be Level 4. It’s also often asserted that Google will make Level 4 vehicles available as early as 2017; however, such reports venture into the realm of hype. The buzz took off in 2012 after one of Google’s founders said of their self-driving car: "you can count on one hand the number of years until ordinary people can experience this." One interpretation of that quote: Google will bring Level 4 vehicles to market by 2017.
In fact, Google has not explicitly promised to deliver a Level 4 vehicle on that timeline, and there are clues that their current focus is on less advanced technology. Last fall, Anthony Levandowski, who leads their self-driving car project, mentioned they’re working on ensuring a human can take control safely when approaching a freeway exit, for example, or when the system is "broken"; he also acknowledged that the car must be able to manoeuvre to a stop if the human fails to take over. Levandowski has also cautioned that "in what form [their technology] gets released is still to be determined." In February of this year, the leader of Google's safety team likewise disavowed any promises about when or in what form the vehicle will come to market. Google publicly aspires to Level 4—the popular video of a vision-impaired man running errands in a test vehicle is one captivating sign—but they are coy on when they might achieve it.
Media coverage can also give an incomplete picture of the state of the art—for example, currently operating automated mining trucks are often mentioned, but it’s less often noted that they drive slowly and often simply stop to avoid obstacles. And while Google’s self-driving cars have racked up around 700 thousand miles in testing, most of those miles have been on freeways, in good weather, under continuous monitoring by at least one human who takes over when conditions look more challenging. Under this vigilant supervision, the Google car can now handle some modestly complex city driving scenarios, but there remain countless, much trickier, more nuanced situations it must still master. Leading experts, such as Steven Shladover of the California Partners for Advanced Transportation Technology, argue that it will take a giant leap to get to systems that reliably and accurately sense their environments in adverse conditions, interpret myriad chaotic and ambiguous situations, predict the behaviour of different road users, and surmount legal, liability, and ethical obstacles—and do it all affordably. In short, Level 4 and driverless taxis could be decades away.
Meanwhile, less advanced, but still impressive, automation technologies are definitely coming soon. Some are already here: for example, some Mercedes vehicles offer simultaneous automatic control of speed and steering (with numerous restrictions). They and other auto manufacturers are developing more capable technologies—though admittedly, the jury is out on whether they will actually make travel easier and safer because of concerns about how humans and machines will share driving responsibilities. And there are already small, light, fully automated vehicles. They aren’t Level 4 because they operate at low speeds, simply stop in risky situations, and are segregated from conventional vehicle traffic. (They are Level 4 in a different classification scheme, developed by the Society of Automotive Engineers.) Such vehicles could serve retirement communities, university campuses, and so on. And automating buses could bring big benefits, such as increased travel speeds and lower labour costs. Even the less advanced technologies can support completely driverless operation when buses operate on busways with adequate protection from other vehicles or hazards.
Once driverless taxis are on the streets, they will play an important role, but primarily in lower-density areas, as a complement to buses and trains. And while the advanced automation they require may be decades away, other automation technologies can improve mobility in the meantime. It's prudent to prepare for the possibility that driverless taxis could arrive in the next few years—but it’s especially critical to start planning to realize the benefits from the less advanced technologies that are already here and those that are definitely emerging soon. They may well be the most advanced automation technologies on the roads for years to come.
Antonio Loro is an urban planner with a special interest in transportation innovations. In his graduate research at McGill University, he investigated the potential impacts of emerging vehicle automation technologies. His research was supported by TransLink in Vancouver and Metrolinx in Toronto. The views expressed in this article are those of the author and do not necessarily represent the views of, and should not be attributed to, TransLink or Metrolinx.