New York City has about 13,000 licensed taxicabs, and sometimes that’s not enough. Ask anyone who has tried hailing one in midtown late on a weekday afternoon.
Often it’s too many; ask anyone who has tried to drive a personal car through the honking yellow swarms, or to get a good night’s sleep in a Manhattan apartment with the window open a half-inch.
A new study from mathematicians at MIT and Cornell suggests that it’s way, way more vehicles than necessary, operating in an antiquated system of gross inefficiency that could be toppled by the disruptive alternative of smartphone hailing pioneered by Uber, Lyft and other services.
If those providers were coordinated by an intelligent ridesharing system, the researchers find, the city could get by on a fleet of just 3,000 four-passenger cars, with an average wait under three minutes between calling a cab and climbing inside. With 10-passenger vans the number could fall to 2,000.
But remaking the Big Apple’s taxi system isn’t their objective. Their focus is on ways of “transforming urban mobility by providing timely and convenient transportation to anybody, anywhere, and anytime. These services have the potential for a tremendous positive impact on personal mobility, pollution, congestion, energy consumption, and thereby quality of life.”
The cost of congestion in the United States alone is roughly $121 billion per year or 1% of GDP, which includes 5.5 billion hours of time lost to sitting in trafﬁc and an extra 2.9 billion gallons of fuel burned. These estimates do not even consider the cost of other potential negative externalities such as the vehicular emissions (greenhouse gas emissions and particulate matter), travel-time uncertainty, and a higher propensity for accidents.
Recently, the large-scale adoption of smart phones and the decrease in cellular communication costs has led to the emergence of a new mode of urban mobility, namely mobility-on-demand (MoD) systems, led by companies such as Uber, Lyft, and Via. These systems are able to provide users with a reliable mode of transportation that is catered to the individual and improves access to mobility to those who are unable to operate a personal vehicle, reducing the waiting times and stress associated with travel.
Not just a New York issue
There are some problems with the paper, which we’ll get to in a moment. And to be sure, the congestion and tailpipe pollution that burden New York are not so serious in, say, Minneapolis or St. Paul, where you could argue that the larger challenges are deficits in conventional public transit (plus a taxi system that has never offered much of an alternative).
But it has been a while since I’ve run across such an intriguing idea for reducing the awful costs of our reliance on the automobile, while preserving or even enhancing its utility in ways that benefit both people and the planet we’re supposed to be tending.
The new study, published last week in the Proceedings of the National Academy of Arts and Sciences, is not the first to look at the potential of Uber/Lyft-enabled ridesharing. (Nor is the ridesharing concept alien to the services themselves, which offer UberPOOL and Lyft Line; in April 2015, when it had been operating for less than a year, Lyft reported that rideshares accounted for more than 30 percent of its trips in New York — and more than 50 percent in San Francisco.)
However, earlier research had significant limitations — like a ceiling of two passengers per trip or, worse, no ability to add passengers to a trip after it had commenced. The MIT/Cornell model allowed both; it used real-time calculations to re-route trips for greater efficiency and also routed idle cars to areas where demand was building.
According to MIT’s announcement of the study, co-author Daniela Rus of MIT’s Computer Science and Artificial Intelligence Laboratory said:
To our knowledge, this is the first time that scientists have been able to experimentally quantify the trade-off between fleet size, capacity, waiting time, travel delay, and operational costs for a range of vehicles, from taxis to vans and shuttles.
And contrary to some reports, it did not base the calculations on driverless vehicles; in fact it considered the interests of drivers as well as riders. Rus again:
Instead of transporting people one at a time, drivers could transport two to four people at once, resulting in fewer trips, in less time, to make the same amount of money. A system like this could allow drivers to work shorter shifts, while also creating less traffic, cleaner air and shorter, less stressful commutes.
3 million actual trips plotted
So how did the modeling work? Researchers began with data on origins and destinations for 3 million actual taxi trips made in New York during the week of May 5, 2013.
The system works by first creating a graph of all of the requests and all of the vehicles. It then creates a second graph of all possible trip combinations, and uses a method called “integer linear programming” to compute the best assignment of vehicles to trips. After cars are assigned, the algorithm can then rebalance the remaining idle vehicles by sending them to higher-demand areas.
You can see a video animation here of how the trips flow under different scenarios of demand, group size and so forth. And you know, from way up here all those taxis look like ants!
Over at Ars Technica, science editor John Timmer added some helpful amplification in plain English:
Figuring out how to ride share is more than just matching passengers with vehicles. You want to match them with a vehicle that’s already nearby, meaning the wait for the ride is minimal. Then, other passengers have to be added in a way that doesn’t result in a significant diversion from the first passenger’s trip. …
A first round of trip assignments is done using what’s called a greedy algorithm, which simply starts with the longest trips first and tries to minimize travel delays. From there, the algorithm attempts optimizations, but it’s capable of producing an answer at any time if a decision has to be made. In addition, it’s possible to make some optimizations like limiting the cars that are considered for servicing a trip (to ones that start off within a set distance of the passenger, for example). That saves considerable computational time.
Speaking of time, the MIT model predicts that the algorithm could dispatch, route and re-route 3,000 vehicles so efficiently that the average wait to be picked up would run 2.8 minutes — which might actually be an improvement over current reality — and the average ride would take only 3.5 minutes longer than a simple point-to-point trip.
As you can imagine, the study has provoked a lot of comment and some criticism, much of centered on Americans’ supposed reluctance to get into a car or van with strangers. A more thoughtful view came from David King, who teaches urban planning at Arizona State and was quoted over at CityLab:
For starters, the researchers assume that each taxi trip in that dataset carries one passenger (the data doesn’t specify this) when in reality, a lot of cabs are already shared by a party of multiple passengers. “Families get in them, or a couple is going out to dinner, so it’s not just a matter of once this vehicle picks up Person 1, it can then pick up Persons 2, 3, 4, and 5,” because the first party may have already filled up half of the seats, he says.
His own research indicates that the average occupancy number is 1.6, which means the 430,000 trips made in one day during that week in 2013 could have carried as many as 688,000 passengers. (For her part, Rus says the parameters of the algorithm can be adjusted and would be tested over time to find that optimal solution.)
Then there’s still the problem with why passengers hail taxis in the first place. “There’s a lot of interest in sharing taxis, and there has been for decades, but nobody ever does it,” King says. “Taxis are a premium service, and the wealthiest use taxis the most because they have a high value of time and they value their privacy. The low-income use taxis when they don’t have any other choice.”
Not just taxi sharing
Maybe, maybe not. Taxi sharing has always been very difficult in New York (and elsewhere) unless all passengers are gathered for the first departure, and even then it’s likely to be greeted grumpily by drivers, who after all would prefer to collect each fare and tip separately. (After quoting King, CityLab’s Linda Poon recounts a short-lived experiment in which New York set up special cab stands for sharing; the cars just blew right by.)
As for premium pricing, it’s not hard to imagine the cost per ride coming down and even way down; a sharing system’s efficiencies would seem to inevitably lower operating costs while raising ridership per driving hour. Competition among providers would likely expand, while the domination of old-line taxi companies would further contract.
Finally, it’s worth remembering that the biggest obstacle to wider use of public transportation, whether in Manhattan or Minneapolis, is generally agreed to be the inconvenience of fitting your trip to the system’s schedule and fixed routes — not the necessity of riding with strangers.
Actually, as a longtime subway and bus commuter, I usually found the company to be an interesting side benefit of going from A to B without a steering wheel in my lap.
* * *
The full paper, “On-demand high-capacity ride-sharing via dynamic trip-vehicle assignment,” can be read here [PDF] without charge.