Industries / Transport and vehicles / Aviation
Scheduling and crews
Airline planners decide which flights to run and when, sell the seats at prices that change with demand, and assign pilots and cabin crew to every flight within legal limits on duty and rest.
6
Milestones achieved, from 1961 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1961
Crew schedules by computer
A crew pairing is a run of flights a crew works out from its home base and back, often with a night away. Murray Spitzer presented a computer method for building low-cost crew schedules at the 1961 symposium of AGIFORS, the airlines' operations research society. By the late 1980s American Airlines built pairings with a program called TRIP, which improves them by solving many small subproblems, and the airline's researchers reported that improvements to it saved over $20 million a year (Anbil and others, 1991).
Credit
Murray Spitzer; American Airlines Decision Technologies, with Ellis L. Johnson of IBM
1964
Seats booked by computer
Before SABRE, American Airlines agents tracked seats on handwritten cards filed around a turntable, and a booking took 90 minutes on average, by IBM's account. SABRE, built by American and IBM after a 1953 meeting between the airline's president and an IBM salesman, cut that to seconds; it was tested in 1960, took over American's reservations in 1964, and handled 7,500 reservations an hour by the mid-1960s. The Computer History Museum notes it was not the first computer reservation system: TWA and other airlines used Teleregister's Magnetronic Reservisor by the late 1950s.
Credit
American Airlines (C. R. Smith) and IBM (R. Blair Smith, Thomas J. Watson Jr.)
1991
Seat prices set by models
Yield management, now called revenue management, decides how many seats on each flight to sell at each fare and how far to overbook. American Airlines' operations research group split the problem into three smaller ones: overbooking, discount allocation and traffic management. American estimated the benefit at $1.4 billion over three years, and the work won the 1991 Franz Edelman Award from INFORMS, the operations research society.
Credit
Barry C. Smith, John F. Leimkuhler and Ross M. Darrow of American Airlines Decision Technologies; Thomas Cook, who led the group from 1982
2011
Rest rules rewritten after crash
After Colgan Air Flight 3407 crashed near Buffalo, New York, on 12 February 2009, the US Federal Aviation Administration (FAA) issued a new pilot duty and rest rule on 21 December 2011. It requires at least 10 hours of rest before each duty period, 2 hours more than before, and raises the minimum time off in any seven days from 24 to 30 hours. Passenger airlines had two years to comply; all-cargo airlines were left out but may opt in. The rule, now 14 CFR Part 117, also lets an airline seek FAA approval for a Fatigue Risk Management System (FRMS), an alternative based on data the airline collects about its own crews' fatigue.
Credit
FAA
2022
Crew recovery collapses
A winter storm disrupted all US airlines in late December 2022, but Southwest Airlines kept failing after the others recovered. By 24 December its ability to reassign crews had broken down, and pilots and flight attendants were stranded along with passengers, the Associated Press reported from the February 2023 Senate hearing; the pilots' union said it had warned about crew scheduling many times since a smaller breakdown in October 2021. Southwest canceled 16,900 flights and stranded more than 2 million passengers, and in December 2023 the US Department of Transportation fined it $140 million.
LessonCrew recovery software must be sized for the worst week of the year, not an average one.
Credit
Southwest Airlines; findings by the US Department of Transportation and the Senate Commerce Committee
2025
New rules, too few pilots
India's aviation regulator, the Directorate General of Civil Aviation (DGCA), brought in stricter Flight Duty Time Limitations (FDTL) in two phases, the second on 1 November 2025, limiting night duty and night landings. Between 3 and 5 December 2025, IndiGo canceled 2,507 flights and delayed 1,852. IndiGo said it had misjudged its crew needs: it needed 2,422 captains and 2,153 first officers for its A320 fleet and had 2,357 captains. The DGCA's inquiry found that IndiGo had over-optimized schedules and crew use without enough buffer, and fined it ₹22.2 crore (₹222 million) in January 2026.
LessonA roster planned to the last legal minute breaks when the rules tighten.
Credit
IndiGo; inquiry by the DGCA
Watch for
Fares priced by AI
Delta Air Lines says it prices part of its domestic network with software from Fetcherr, an Israeli firm, that it calls "AI-enhanced pricing". Delta's president, Glen Hauenstein, said on 10 July 2025 that about 3% of domestic fares were set this way and the goal was about 20% by the end of 2025, while calling it "a heavy testing phase". US Senator Ruben Gallego called it "predatory pricing".
Achieved whenDelta reports the share reached, and an outside party, such as a regulator or an academic study of filed fares, measures the effect on prices and revenue.
Claimed by
Delta Air Lines, July 2025
Recovery plans in minutes
SITA, the airline-owned IT company, bought Big Blue Analytics, maker of a disruption tool called OCCam, in June 2026. SITA says OCCam weighs aircraft, crew, passenger and maintenance limits together, produces a single recovery plan "in minutes", and can "cut recovery costs by up to 30 percent". No airline user or measured result is named.
Achieved whena named airline or an independent party publishes measured recovery times and costs before and after use.
Claimed by
SITA, June 2026
Open
01
Pilots needed under new rules
An open calculator that takes a public flight schedule and a rule set, such as India's FDTL before and after November 2025, FAA Part 117 or the European Union's rules, and estimates how many captains and first officers it takes to fly that schedule legally, with a stated allowance for sickness, leave and training. Test it on IndiGo's A320 schedule for November 2025, rebuilt from public flight-tracking data.
The briefCredit · unclaimed02
Replay a meltdown in public
A public benchmark for crew recovery: a set of disruption cases, each with a flight schedule, aircraft rotations, crew pairings, duty rules and a scoring program that counts canceled flights, stranded crews and rule breaches. Include one case rebuilt from Southwest's flights of 21 to 31 December 2022, with crew pairings made up to match the published schedule, since real crew records are private.
The briefCredit · unclaimed
Open 01 · the brief
Pilots needed under new rules
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open calculator that takes a public flight schedule and a rule set, such as India's FDTL before and after November 2025, FAA Part 117 or the European Union's rules, and estimates how many captains and first officers it takes to fly that schedule legally, with a stated allowance for sickness, leave and training. Test it on IndiGo's A320 schedule for November 2025, rebuilt from public flight-tracking data.
Why it is still open
IndiGo said it misjudged its own crew needs under the new rules, and the DGCA's inquiry found its planning did not account for them. The crew planning tools airlines use, such as Lufthansa Systems' NetLine/Crew, are closed products. Public code on GitHub includes calculators that check one pilot's duty day and a few crew-pairing solvers, but this research found no public tool that estimates a whole airline's crew need under competing rule sets.
Built with what exists
Flight records from the OpenSky Network, a volunteer receiver network whose data is free for research and non-commercial use; an open solver such as HiGHS or Google OR-Tools; the open crewml package for building pairings; a language model to draft the rule clauses as code from the regulations' text, with a reviewer pass that flags doubtful clauses for a person to check (see the Note "Seven more flagged").
First step
Write the DGCA's weekly-rest and night-duty clauses as ten test cases with known answers, and pull one week of IndiGo departures from OpenSky.
Done looks like
A published estimate of IndiGo's A320 crew need under the old and new Indian rules, set against IndiGo's own figures of 2,422 captains and 2,153 first officers, with the gap stated as a percentage, and the same run repeated for two more airlines, with code and inputs public.
Open 02 · the brief
Replay a meltdown in public
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public benchmark for crew recovery: a set of disruption cases, each with a flight schedule, aircraft rotations, crew pairings, duty rules and a scoring program that counts canceled flights, stranded crews and rule breaches. Include one case rebuilt from Southwest's flights of 21 to 31 December 2022, with crew pairings made up to match the published schedule, since real crew records are private.
Why it is still open
The public test set most used for airline disruption research, the 2009 ROADEF/EURO Challenge set from Amadeus, leaves crews out "in order to limit the scope of the problem", and a 2026 study of integrated recovery by Lin and colleagues also covers only flights, aircraft and passengers. A 2025 review by Rodrigues and colleagues says practice "often relies heavily on manual adjustments" and that few studies treat aircraft and crew recovery together. A few small projects on GitHub simulate crew recovery, but none publishes a scored benchmark built on a real event.
Built with what exists
The US Bureau of Transportation Statistics on-time data, which lists every US airline flight with its tail number and whether it was canceled; the ROADEF 2009 file format as a starting point; HiGHS or Google OR-Tools for a baseline solver; Python.
First step
Download the December 2022 Southwest records from the Bureau of Transportation Statistics and chart cancellations by day and by airport.
Done looks like
At least 20 public cases with a scoring program that anyone can run, baseline results for each, and for the Southwest case a published count of how many of the canceled flights the baseline could have kept flying under stated assumptions.