In scheduled air transportation, airline profitability is critically influenced by the airlines ability to: 1) estimate passenger demands; 2) construct

profitable flight schedules (referred to as the airline schedule planning process); and 3) determine the fare levels for a set of products in an origin-destination market (referred to as the pricing process) and determine how many seats to make available at each fare level (referred to as revenue management).

Schedule planning is a multi-step process, usually separated into four, sequentially solved sub problems: schedule design, fleet assignment, maintenance routing, and crew scheduling

Aircraft routing Airlines must meet maintenance requirements in making aircraft rotations. In order to ensure air travel safety, the Federal Aviation Administration (FAA) requires airlines to perform four types of aircraft maintenance called A-, B-, C-, and D-checks. Airlines have self-imposed inspection and maintenance requirements which are more stringent than FAA requirements. The four types of maintenance vary in scope, duration of time, and frequency. Among them, C-, and D-check requirements do not affect daily scheduling and routing since the aircraft is taken out of the schedule. A-checks involve a routine visual inspection of all major systems which require 4 hours. It is common for airlines to perform A-checks in every 35 hours of flying. B-checks include a thorough visual inspection plus lubrication of all moving parts and are performed every 300 to 600 flying hours. The B-checks require a 10 to 15 hour long stay at the maintenance hangar. (Clarke et al. 1996) incorporates B-check requirements into the fleet assignment problem, and (Clarke et al. 1997 incorporates A-check requirements into the aircraft rotation problem. Aircraft rotation is the routing of individual airplanes, which is determined after fleet assignment. Typically, the airlines require that each aircraft in a fleet fly an identical route consisting of all the legs assigned to that fleet, which implies that the route must be a cycle.

When the flights covered by aircraft are separate from the flights covered by another aircraft in the fleet, it is called a broken rotation. A locked rotation is a broken rotation where the separate sequences do not have any locations in common (Clarke et al. 1997). Locked rotations are unacceptable to almost all airlines. Airline business processes related to tactical planning consist of schedule planning, fleeting, aircraft rotation, and crew pairing .In schedule planning; a set of flights with specific departure and arrival times is constructed. Next is fleeting, which assigns an equipment type (such as Airbus 320, Boeing 737-500, etc.) to each individual flight. The objective of the fleet assignment model (FAM) is to maximize profit subject to the number of available aircraft and other operational constraints. The problems that follow decompose based on the fleeting solution, i.e. there is a separate problem for every equipment type. The aircraft rotation problem or aircraft routing is to find a set of generic aircraft routes that satisfy maintenance requirements. The crew pairing optimization follows. In crew pairing a set of crew itineraries or pairings is constructed. The goal is to minimize the crew cost and each flight must be covered by exactly one pairing. A few weeks before the day of operations, the actual tail numbers are assigned to each flight and monthly crew rosters or bid lines are assigned to every individual crewmember. Throughout tactical planning, revenue management related processes match demand with supply, which is the sit capacity of a given equipment type. At present, each one of these problems is modelled independently of the remaining problems. (The only exception is a very recent research work focusing on combining selected two consecutive stages.) The models are solved sequentially and the output of one stage is the input to the next stage. Clearly there is interdependency between the various stages. For example, the fleeting solution decomposes the problems that follow by equipment type. The aircraft rotation and crew pairing problems are then solved over a subset of flights pertaining to a single equipment type. This is due to the fact that each cockpit crew is qualified to fly a particular set of equipment types. But this dependency of crews on equipment types is not captured in fleeting. In the aircraft routing stage decisions are made without considering the impact on the quality of the crew pairing solution. The interaction between these two stages is in the fact that a crew can have a connection shorter than a predefined number only if it stays on the same aircraft. Solutions obtained by using this sequential methodology can therefore be suboptimal. Ideally all the tactical planning problems should be solved as a single large-scale problem. Abstract: In many ways a major airline can be viewed as one large planning problem which is usually approached as many interdependent smaller (but still imposing) planning problems. The list of things which need planning seems endless: crews, reservation agents, luggage, flights, through trips, maintenance, gates, inventory, equipment purchases. Each planning problem has its own considerations, its own complexities, its own set of time horizons, its own objectives, but all are interrelated. In this talk, we will briefly look at a few of these airline planning problems. For each, we will outline the basic problem, considerations, and objectives. In addition, for each planning problem we review, we will discuss how one might quantitatively approach the problem so as to intelligently support the planning process. At the IMA Industrial Problems Seminar on January 30, 1998, Richard Stone reported on airline planning problems. In many ways a major airline can be viewed as one large planning

problem that is usually approached as many interdependent smaller but still imposing planning problems. The list of things that need planning seems endless: crews, reservation agents, luggage, flights, through trips, maintenance, gates, inventory, equipment purchases. Each planning problem has its own considerations, its own complexities, its own set of time horizons, its own objectives, but all are interrelated. In this talk Dr. Stone looked briefly at a few of these airline planning problems, outlining for each the basic problem, considerations and objectives. In each case he discussed how one might quantitatively approach the problem so as to support intelligently the planning process. It takes hundreds of people many months to make an airline schedule. The following can be thought of as a check list of the high-level issues to be decided in building an airline schedule.

Markets.

To which airports should the airline fly?

Frequency.

For each airport, how much service should the airline provide?
Hubs.

Which airports should be used as hubs? A hub is an airport used as a central connecting point for traffic.
Fleet

Planning.

Which types of aircraft should the airline purchase? How many of each type?
Banks.

To maximize the number of possible connections and, hence, the number of "city pairs" it serves, an airline tends to organize the hub flights into banks. A bank consists of several dozen planes which: (1) all arrive close in time, (2) pause to let passengers change planes and then (3) all depart close in time. Exactly how many banks there should be, when they should occur, and how they should be structured, are important issues involved in designing an airline schedule.
Timing.

Exactly when during the day should the flights serving the various airports arrive and depart?
Block

Times.

The block time is the time between when a plane leaves the gate at the departure airport and enters the gate at the arrival airport. Since many factors can delay a flight, the larger the scheduled block time, the more likely that the flight will arrive ontime. However, larger block

times also mean fewer flights can be scheduled and higher costs per flight. Achieving an appropriate balance is another issue involved in designing an airline schedule.
Fleet

Assignment.

Which type of plane should be used for each leg? (A leg is a nonstop flight.) Clearly, we would want the larger planes to service the legs with the larger demands. This requires forecasting that future demand. It also requires assigning fleets to legs that best cover the demand given the operational constraints of the schedule.
Through

Flights.

When a plane is part of a bank, it will service one of the arrivals into the bank and one of the departures out of the bank. A passenger traveling on both that arriving flight and departing flight does not have to change planes to make the connection. This type of connection is called a direct flight or through flight. Passengers tend to prefer direct flights rather than regular connections. Therefore, planning must be done, given the arrivals, departures, and other considerations, so as to create those through flights that would attract the most passengers.
Planned

Maintenance.

The schedule must ensure that, on a regular basis, each plane stays overnight at a maintenance base for a minimum period of time.
Meal

Service.

The level of meal service must be decided for each flight.

Crew

Planning.

Pilots and flight attendants must be scheduled in a manner that meets federal, airline, and union regulations, in addition to ensuring that each flight can be flown.
Gate

Plotting.

At an airline's hubs, when many planes will be on the ground at once, the problem of parking all the planes simultaneously can be quite difficult. There are several physical and operational considerations to take into account in addition to such goals such as minimizing the distance passengers and luggage must travel within the airport.
Station

Manpower.

This involves the scheduling of the many people working at the airports. This includes both the customer service personnel helping passengers in the terminal as well as ground service personnel handling the aircraft on the ground.
Recovery.

As with all things, even if all the above planning is done as well as possible, many factors cause deviations from the plan. Recovery is the day-to-day concern of returning to plan as smoothly as possible. Some of these problems are not mathematically tractable. For a more detailed analysis Dr. Stone selected one of these problems that is tractable mathematically: Crew Planning Compared to most of the other planning problems listed, crew planning is relatively easy to model mathematically and interpret as an optimization problem. Still, efficiently solving the model and producing a practical crew plan can be quite complicated. In the following, we will be concerned with the flight crew (pilots), but similar issues arise when dealing with the cabin crew (flight attendants). Airline crews do not follow normal 9-to-5 work days. The actual work day for a crew member can start at any time during the day (or night) and can be somewhat longer or shorter than 8 hours. Such a work day is called a DUTY PERIOD and consists of briefings, flight legs, ground time, and debriefings. A work week, that is a sequence of duty periods separated by overnight stays, is called a TRIP or PAIRING. A trip must start and end at the crew member's home base. There are many federal, airlines, and union rules regarding duty periods and trips. These dictate the maximum amount of time within a duty period that a pilot can spend flying a plane. These, also, dictate the minimum amount of time that overnights between duty periods can be depending on the length of previous overnights, the amount of flying a pilot has done, and the type of flying that was done, e.g., flying a red-eye flight requires more rest time. Normally pilots are paid for the total block time (see above) of the flights that they flew. However, the rules dictate that a pilot must be paid a certain amount over and above this total block time depending on how the duty periods and trips are structured. A flight leg that a pilot flies is called a LIVE flight for that pilot. When pilots fly on legs as passengers during a duty period, so as to get to the next live flight, they are said to DEADHEAD on that leg. The rules can dictate that the pilots must be paid for DEADHEAD flights. Also, the rules can dictate that pilots must be paid a certain minimum for duty periods and/or trips, and this minimum can be an absolute amount and/or a percentage of the total time of the duty period and/or trip. This extra amount that pilots are paid above the total block time is called the PENALTY and represents the inefficiencies contained in the crew plan. The object of crew planning is to design a set of trips that follow all the rules, ensure that each leg in the schedule has a crew to fly it, and minimizes the penalty. We say that a trip is LEGAL if it satisfies all the various rules. A step towards modeling the crew planning problem would be to define a binary matrix A whose rows represent the flight legs in the schedule and whose columns represent all possible legal trips. The element Aij would equal 1 if flight leg i is a live flight for trip j. Otherwise, Aij would equal 0. We next determine a vector c such that cj is the total crew cost associated with trip j. The crew planning problem can now be stated mathematically as: