Autonomous taxis: Why you may never own a self-driving car

As the once unimaginable self-driving car moves closer to becoming a reality, the next question is “When can I buy one?” At the same time, some researchers, like Princeton’s Alain Kornhauser, and the University of Texas’s Kara Kockelman, have started to wonder whether you’ll ever need to. They envision fleets of autonomous vehicles that will combine the convenience of not having to drive yourself with the flexibility of a scaled-up and always available Uber-like taxi service — and without the cost of hired drivers. Since we first wrote about this topic, the landscape has changed dramatically for the better, both for autonomous vehicles, and for ride-hailing and ride-sharing services. They’re also key to future smart cities, which we’ll be covering all this week here at ExtremeTech in our first-ever Smart Cities Week.

Instead of seeming like science fiction, self-driving cars are now on the roadmap for not only nearly every car maker, but for ride-hailing giant Uber. It’s easy to write off Uber’s investment as hubris, but there has been an increasing amount of research that shows how fleets of shared autonomous vehicles would make good economic sense. One recent study done by researchers at the University of Virginia and the University of Texas, using Austin as a model, estimates that over 1/4 of what would previously have been trips in private vehicles would move to a shared fleet if it was priced around $1/mile. Research firm ARK Invest concludes that the cost of operating a shared autonomous vehicle fleet could be as low as $.35 per mile, less than 1/10th as much as the cost of traditional taxis, and about half of owning a car.

For most of us the idea sounds pretty far fetched. After all, what about peak times like rush hour when everyone seems to want to go somewhere at the same time? Of course, that’s just our intuition, not science. Since we originally wrote about the work Kornhauser and his students had done building and refining a realistic model of actual travel needs and car usage using auto trip data from New Jersey, another resource has become available — a complete database of New York City taxi trips. Researchers are using that data as a proxy for overall transportation demand, and modeling how various types of ridesharing service options could reduce vehicle miles.

New Jersey has a huge variety of population densities as shown by the team's model -- making it a good simulation for most of the USPrinceton’s New Jersey model started with 2010 census data and built on it with information from other behavioral studies and surveys about where people live, work, and travel. It includes data for the 430,000 businesses and 18,000 schools in the state, along with the 120,000 blocks of census data. Each person (including residents and about 500,000 commuters from out of state) was assigned a place of work or school, as appropriate, and then heuristics were used to model the trips each would take on a typical day. The result was a massive model of the over 30 million vehicle trips taken on a given day in New Jersey, including their timing, origin, and destination — down to the street address. Currently there are about 4 million private vehicles serving its population and providing those trips. A similar model for the entire US would need to include about 1.2 billion trips taken by the over 300 million US residents on a given day. Then the team overlaid a fleet of autonomous taxis that they call aTaxis, and looked at how well it could do the job instead.

The New York City data covers every single taxi trip taken in 2013, including time and start and destination. To simplify his analysis of ridesharing options, Princeton’s AJ Swoboda divided the city into .1-mile by .1-mile “pixels” and assumed patrons would be willing to share rides with those needing to travel from the same pixel at a similar time to a destination that made a combined trip make sense. The study makes use of another major change over the last two years. Hailing a vehicle through a mobile application is now commonplace. The use of that type of technology, which can in turn be linked to a smart back-end that can optimize vehicle dispatch and routing, makes ridesharing a lot more practical than the old-fashioned “whistle for a taxi” hailing model most of us grew up with.

You can easily see where the demand for taxis is highest in this visualization generated by Princeton's SwobodaThe team’s vision for aTaxis is as a less-expensive, more-convenient service that combined all the good qualities of Zipcar and Uber, without the hassle of having to drive yourself or of paying for a driver. Obviously the whole idea hinges on truly autonomous vehicles — often called Level 4 automation — so it isn’t going to happen quickly. It also depends on enough usage to make the capital investment worthwhile. Kornhauser is predictably optimistic about the potential demand, as he puts it, “you get to buy mobility by the drink, rather than by the bottle.”

One way to jumpstart the fleet would be a system similar to the one used by RelayRides, where individuals own each car. Owners could simply mark their cars as available and have them drive off to ferry other passengers and earn them some rental revenue — this could be particularly popular with commuters and students who leave their cars sitting all day. Even with that kind of a boost, aTaxi systems are likely to start in heavily populated areas and only slowly spread out from there. Austin seems to be a favorite candidate on the part of researchers. It makes sense as an area with awful traffic that hosts the University of Texas, and is also a test site for self-driving cars including Google’s.

Originally, it was thought that the biggest benefit of sharing cars would be cost savings. But as the rapid growth of Uber and Lyft shows, the convenience of on-demand transportation and the lower hassle of a “no-car” lifestyle is very appealing, especially to Millennials. This will only accelerate the move to shared autonomous vehicle fleets once the technology becomes available. Cost savings, though, are still a major driver for their creation. Overall, it is estimated that a shared autonomous vehicle fleet could serve travelers’ needs with only half the number of cars on the road today. In addition to saving vehicle costs, parking needs and vehicle congestion would be greatly reduced. US cities have parking areas that if laid out flat would cover from 20% to 80% of their land area. Much of that could be reclaimed for additional housing, parks, or businesses. Congestion would be reduced through greater ride sharing and easier access to mass transit.

Even though declining there are still over 12 thousand bus-related injuries in the US each yearLow-use bus routes are perfect candidates for replacement by an aTaxi service. Replacing them would provide benefits in lower cost, less pollution, and improved safety. As retired New Jersey transit planner Jerome Lutin puts it, “If you can’t get more than 10 people on a bus, or five people on a bus, then why bother running it? You’re wasting diesel fuel.” As far as safety, while bus drivers are trained and licensed professionals (in the US at least) they can suffer from the same issues of fatigue and distractions as any driver. In New Jersey alone, the damage and liability costs of bus-related accidents are nearly $500 million each year. aTaxis — like all eventual self-driving cars — are expected to be significantly safer on average than their human-piloted counterparts.

Here too, the advent of mobile apps can play a large role. While most current apps, like those for Uber, Lyft, or taxi fleets, only connect you with one vehicle and one ride, newer versions are emerging that will allow users to plan and ticket trips that include both a car and some form of mass transit. Just recently, Mercedes subsidiary Moovel announced that it had acquired multi-modal ride-planning companies GlobeSherpa and RideScout. No doubt it will work to tie them into its car-sharing offering Car2Go.

Even the classic commuter rail lines would benefit from a flourishing autonomous taxi system. One of the big problems with commuter rail today is the hassle and expense of parking at the station. Acres of prime real estate are tied up with massive lots all along the New Jersey transit rail lines, and most of them are packed full before mid-morning. Those cars sit idle all day, while if they were aTaxis they could be productively shuttling students to school or shoppers to stores. Another benefit of more convenient access to rail lines should be an increase in their use — further helping reduce carbon emissions from commuter cars and reducing congestion during rush hour.

A typical aTaxi trip is modeled -- in this case assuming a short walk to an aTaxi standThe Achilles heel of any taxi system is peak time. Whether it is rush hour or simply an impromptu downpour, we’ve all experienced the frustration of a sudden lack of available cabs. Similarly, car sharing services like Zipcar are only helpful with some advance planning, as cars are likely to be sold out during popular periods. Simply making the cars autonomous doesn’t solve this problem. However, a cloud-based dispatch infrastructure, similar to that used by services like Uber today, would provide an intriguing solution if customers are willing to share. Uber itself has clearly figured this out, and is making its own large investment in autonomous vehicles.

An analysis done by the aTaxi project showed that ridesharing potential is fortunately highest exactly when demand is peaking — with many commuters leaving the same places of work at the same time, for example. Clearly even this fairly simple ride sharing is a loss of apparent convenience compared with having your own private vehicle, but for many that will be offset by the freedom from parking and maintaining your own vehicle. For the system to achieve real savings in the number of cars on the road, users will need to be willing to do some impromptu sharing — perhaps with other passengers being picked up and dropped off by the aTaxi while it is en route. For their analysis the team assumed that sharing couldn’t add more than 20% to the total time spent on a trip.

No matter how attractive sharing autonomous vehicles may become financially, some drivers will simply want to own their own car — because they want immediate access, or they want to leave belongings in the car, or they want to customize it, or they just enjoy driving. For those folks, self-driving will just be another feature from the land of high-tech wizardry that will give them a chance to snooze on boring highway stretches, or get themselves home safely after over-indulging at a party. But for a large number of car owners — and especially potential car owners — the future may wind up consisting of shared access to a well-maintained fleet of self-driving cars.

To learn more about aTaxis, and other smart driving solutions, hit up Kornhauser’s Smart Driving Cars website, or UT’s Center for Transportation Research

Leave a Reply

Captcha image