~ .

~
One small part of an airline's schedulingproblemsis disc~sed below.
Large airlines Cacea staffingproblem known as fue crew schedulingproblem. The
requirementsto be coveredare the crew requirementsof fue flights which the line is
committed to fly during the next schedulingperiod, e.g., one month. A specific crew
during its working day typically fly a numberof flights, usually, but not nect:ssarilyon
the sameaircraft. The problem is to determinewhich flights should comprisethe day's
work of a crew.
The approachtakenby a numberof airlines is similar to fue approachdescribedfor
staffing problems:(1) Identify the demandrequirements,i.e., the flights to be covered;
(2) Generatea large numberof feasiblecollectionsof flights that one crew could cover
in a work period; (3) Selecta minimum cost subsetof fue collectionsgeneratedin (2) so
that the cost is minimized and every flight is containedin exactly one of fue selected
collections.
LP can be used for step (3). Until recently,most large airlines used computerized
ad hoc or heuristic proceduresfor solving (3) becausefue resultingLP tendsto be large
and difficult to solve.Marsten,Muller and Killion (1979),however,describean LP based
solution procedurewhich has been used very successfullyby Flying Tiger Airlines.
Flying Tiger has a smaller fleet than the big passengercarriers, so the resulting LP can
be economicallysolvedandgivesmarkedlylower costsolutionsthanfue ad hoc, heuristic
methods.Theseoptimizing methodsare now being extendedto large airlines.
A drastically simplified version of fue crew schedulingproblem is given in the
following example.This examplehasonly tenflights to be covered.By contrastin 1990,
United Airlines had close to 2000 flights per day to be covered.
Example: Sayre-PriorsCrew Scheduling
The Sayre-PriorsAirline and StormdoorCompanyis a srpall diversified companythat
operatesfue following set of scheduledflights:

Flight Number Origin Destination Time of Day

101 Chicago Los Angeles Aftemoon
410 New York Chicago Aftemoon
220 New York Miami Night
17 Miami Chicago Moming
7 Los Angeles Chicago Aftemoon
13 Chicago New York Night
11 Miami NewYork Moming
19 Chicago Miami Night
23 Los Angeles Miami Night
3 Miami Los Angeles Aftemoon

The flight scheduleis illustratedgraphically in Figure 6.2.

The Flight OperationsStaffwould like to setup a low costcrewassignmentschedule.
The basic problem is to determinefue next flight, if any, that a crew operatesafter it
completesone flight. A basic conceptneededin understandingthis problem is that of a
tour. The characteristicsof a tour are as follows:
E'~'" CHAPTER
6: Covering,
Stalfing,& CuttingStockModels 8

Figure6.2
Flight Schedule

,,.....
,, .....
,, ....
,, ...
,, .
, ..
, ..

Afternoon
,
Night
Morning

. A tour consists of from 1 to 3 connecting flights.

": . A tour has a cost of $2,000 if it terminates in its city of origino
. A tour which requires "deadheading," i.e., terminates in a city other than the
origin city, costs $3,000.
In practice fue calcu1ation of fue cost of a tour is substantial1y more comp1icated
than above. In air1ine par1ancea tour is frequent1y cal1eda "pairing" or a "rotation." The
fol10wing are examples of acceptable tours:

Tour Cost
17, 101,23 $2,000
"B 220, 17, 101 $3,000
i! 410, 13 $2,000

Solving the Soyre/Priors Crew Scheduling Problem

The first thing to do for this small problem is to enumerate all feasib1etours. We do not
considera collection of flights which invo1vean intermediatelayover a tour. There are
10 one-flight tours, 14 two-flight tours, and either 37 or 41 three-flight tours depending
upon whetherone distinguishesfue origin city on a nondeadheading tour. Thesetours are
indicatedin Tab1e6.3.
 Table6.3
List of Tours
One-FlIghl
Tours Cosl Two-FlIghl
Tours Cos! Three-FlIghl
Tours Cos!
1 101 $3000 11. 101,23 $3,000 25. 101,23,17 $2,000
2. 410 $3000 12. 410,13 $2,000 26. 101,23,11 $3,000
3. 220 $3000 13. 410,19 $3,000 27. 410,19,17 $3,000
4. 17 $3000 14. 220,17 $3,000 28. 410,19,11 $2,000
5. 7 $3000 15. 220,11 $2,000 29. 220,17,101 $3,000
6. 13 $3000 16. 17,101 $3,000 30. 220,11,410 $3,000
7. 11 $3000 17. 7,13 $3,000 25. 17,101,23 $2,000
8. 19 $3000 18. 7,19 $3,000 31. 7,19,17 $3,000
9. 23 $3000 19. 11,410 $3,000 32. 7,19,11 $3,000
10. 3 $3000 20. 19,17 $2,000 33. 11,417,13 $3,000
21. 19,11 $3,000 28. 11,410,19 $2,000
22. 23,17 $3,000 34. 19,17,101 $3,000
23. 23,11 $3,000 28. 19,11,410 $2,000
24. 3,23 $2,000 25. 23,17,101 $2,000
35. 23,11,410 $3,000
36. 3,23,17 $3,000
37. 3,23,11 $3,000

Define fue decisionvariables:

Ti = { O1 ifif tour
tour ii is
is used
not usedfor i = 1,2, . . . , 37.

We do not distinguish fue city of origin on nondeadheading three-ftight tours. The forrnu-
lation is (measuring cost in $1,OOOs):

MIN :3 TI +:3 T2 +:3 T3 +:3 T4 +:3 T5 +:3 T6 +:3 T7 +:3 T8 +:3 T9

+:3 TIO +:3 TII + 2 TI2 +:3 TI3 +:3 TI4 + 2 TI5 +:3 TI6 +:3 TI7 +:3 TI8
+ :3 TI9 + 2 T20 + :3 T2I + :3 T22 + :3 T2:3 + 2 T24 + 2 T25 + :3 T26 + 2 T27
+ 2 T28 +:3 T29 + 3 T:30 + :3 T:3I + 3 T32 + 3 T:33 + 3 T:34 +:3 T:35 + :3 T:36
+:3T37
SUBJECTTO
FIOI) TI + TII + TI6 + T25 + T26 + T29 + T34 = I
F410) T2 + TI2 + TI:3 + TI9 + T27 + T28 + T30 + T:3:3+ T35 = I
F220) T3 + TI4 + TI5 + T29 + T:30 = I
F17) T4 + TI6 + T20 + T22 + T25 + T27 + T29 + T:3I + T34 + T:36 = I
F17) T5 + TI7 + TI8 + T:3I + T:32 = I
FI:3) T6 + TI2 + TI7 + T:3:3= I
FII) T7 + TI5 + TI9 + T2I + T2:3 + T26 + T28 + T:30 + T:32 + T:33 +
+T35 + T37= I
F19) T8 + TI:3 + TI8 + T20 + T2I + T27 + T28 + T:3I + T:32 + T34 = I
F2:3) T9 + TII + T22 + T2:3 + T24 + T25 + T26 + T35 + T:36 + T:37 = I
F:3) TIO + T24 + T:36 + T37 = I
END

A PICTURE of fue coefficients gives a better feel for fue structure of fue problem and
is given below. The first constraint, for example, forces exactly one of the tours which
includes ftight 101 to be chosen. ~