Air taxis: Changing the way we fly looks at how “two aging computer geeks are setting out to reinvent business travel” via DayJet, an airline of tiny Eclipse 500 jets — which only hold five people, two pilots and three passengers — offering short flights in Florida:
Florida is an ideal testing ground for this type of service because of its good flying weather, the density and high income of its population, and the miserable commercial air routes within the state.
This can only work with some amazing Operations Research:
How much a DayJet flight costs will depend on how flexible a traveler’s schedule is. The price of each flight will range from $1 to $3 per mile. A 329-mile, one-hour flight between Boca Raton and Tallahassee, for example, will cost nearly $1,000 each way if the traveler can’t give DayJet more than a 75-minute window to work with.But if the customer agrees to fly anytime between, say, 7 a.m. and 1 p.m. — a six-hour window — the flight might cost only $329 each way. The wider the window, the more options DayJet has to meet the reservation.
Behind this reservation system is really complicated mathematics. It’s basically a resource-allocation problem. Given a certain number of planes, routes, and existing reservations, what is the optimal way to reconfigure DayJet’s network of air taxis to accommodate each new reservation request?
“We’ll have to evaluate billions of options and come back to you with a yes or no answer in five seconds,” Iacobucci says. Even a supercomputer would have trouble doing that.
Instead of a supercomputer, Iacobucci has two Russian mathematicians, Eugene Taits and Alex Khmelnitsky, stashed in a windowless room down the hall working on an algorithm they believe will solve the problem. At DayJet, everyone calls them the rocket scientists. Their algorithm quickly creates a best guess as to whether DayJet can meet a request and at what cost. As long as it comes up with an acceptable answer, it can offer a quote.
From the time a quote is given until just before the flight plans need to be filed, DayJet’s system can keep trying to come up with an even better answer that lowers the total cost of the air-taxi network. On rare occasions, that might mean three different planes taking three passengers to the same place, if that’s more efficient for the overall network.
To test this algorithm, Iacobucci is working with some operations research scientists at Georgia Tech who do have access to a supercomputer. It takes them 24 hours to come up with the same answers DayJet’s optimization algorithm comes up with in a few seconds.
In another office farther down the hall (these have windows), Iacobucci keeps his ant farmers. They are complexity scientists, originally from the Santa Fe Institute, who have created a massive simulation of the entire U.S. transportation system. They’ve mapped travel patterns into 10-square-mile blocks, complete with income levels, demographics, historical driving patterns, airport drive times, and airline schedules and fares.
“It’s like Sim City on steroids,” Iacobucci says. After calibrating the simulation to match current travel behavior, the ant farmers introduce DayJet service in different cities and see how the simulated people react. That’s how the DayPorts are selected.