FLIGHT TRANSPORTATION LABORATORY REPORT R 91-4 PRESENTATIONS FROM THE MIT/INDUSTRY COOPERATIVE RESEARCH PROGRAM ANNUAL MEETING, 1991 DEPARTMENT OF AERONAUTICS Belobaba, Williamson, et al. ASTRONAUTICS May 1991 FLIGHT TRANSPORTATION LABORATORY Cambridge, Mass. 02139 FTL REPORT R 91-4 PRESENTATIONS FROM THE MIT/INDUSTRY COOPERATIVE RESEARCH PROGRAM ANNUAL MEETING, 1991 TABLE OF CONTENTS Origin-Destination/Segment Seat Inventory Control: Modeling and Implementation Issues. Peter Belobaba 1 Application of Network Solutions to Origin-Destination Seat Inventory Control. 14 Elizabeth Williamson Planning and Scheduling Tasks in a Dynamic Environment. Lyman Hazelton 46 Concentration in U.S. Air Transportation: An Analysis of Origin-Destination Markets since Deregulation. 73 Jan van Acker Pricing in the Airline Industry: Current Practice and Future Research. Theodore C. Botimer 102 Changes in Origin-Destination Passenger Traffic Flows: Newark Airport. Chung Mak and Peter Belobaba 129 Airline Seat Inventory Control for Group Passenger Demand. Peter Belobaba and Tom Svrcek 157 Airline Crew Scheduling Re-visited. Robert W. Simpson 173 Scheduling Systems: Computer Aids for Execution Rescheduling. Dennis F.X. Mathaisel 187 -1- O-D / SEGMENT Modeling SEAT INVENTORY and CONTROL: Implementation Issues Professor Peter P. Belobaba MIT Flight Transportation Laboratory MIT/Industry Cooperative Research Annual Meeting May 23, 1991 Program -2-2- PASSENGER e ITINERAR Y CONTROL In contra! to the control of seat inventories by flight leg/booking class, passenger itinerary control requires methods that can distinguish among Dassenaer itineraries itineraries vying for seats on the same flight leg, even within the same fare product "category" Passenger Itinerary Control concepts can be applied through: DUAL/OVERLAP FLIGHTS: Two or more flight numbers representing different itinerary "paths" a ssigned to a single flight leg. POINT OF SALE CONTROL: Seat availability is differentiated between points of sale due to currency or net revenue differences. SEGMENT CONTROL: Seat availability managed by booking class and passenger itinerary on a multiple-leg flight with the same flight number. O-D CONTROL: Seat availability managed by fare category and O-D itinerary, even across connecting flights. -3- PASSENGER )VERLAP DUAL / D7W ITINERARY CONTROL FLIGHTS: NWoL09 L4W tW1b34l POINT OF SALE: KL 016 AMs RB SEGMENT CONTROL: CONNECTING O-D AC 113 CONTROL: 2 0 DL145 liou 00s SAV -4- NETWORK OPTIMIZATION APPROACHES Traditional O.R. approach to dealing with itinerary control is to perform a joint optimization over the entire network of flight legs defined by the problem. . Network formulations and/or mathematical programming approaches are used to find the optimal a Illocation of seats to each origin-destination itinerary and fare type (ODF) on each flight leg: -- demand forecasts and fare values for each ODF requires representation can be deterministic or probabilistic -- problem -- solution -- ODF allocations are "optimal" given ensures balanced ODF allocations across flight legs assumed mathematical formulation of problem -5- 9 Implementation of these "optimal" network ODF seat allocations as ODF booking limits, however, can have substantial negative revenue impacts. -- use of partitioned or discrete booking limits lowers expected revenue relative to nested limits revenue impact becomes larger with more ODF allocations -- negative -- dynamic simulations show that use of "optimal" partitioned segment/class limits on a 3-leg flight can result in revenue reductions of 1% to 2% relative to simpleh leg/booking class control The oDtimal solution to an assumed mathematical formulation does not maximize revenues when implemented this way. -6- NESTING OF NETWORK SOLUTIONS OPTIMAL Nestina of ODF Allocations Within Each O-D/Seament 0 Network ODF allocations for each O-D/segment are nested into a shared inventory of seats, in order of normal nested booking classes -- each ODF allocation is treated as a fare class "protectic n level" within the O-D/segment "nest" -- each flight leg still has a discrete allocation of s eats for each segment/O-D itinerary -- summing discrete allocations gives sub-optimal nested booking limits within each O-D/segment nest simply -7- EXAMPLE: 2 segments on Leg A-B (Capacity = 100) ODF Optimal Allocation Segment Nesting YAB 8 42 BAB 12 34 MAB 06 22 QAB 16 16 YAC 18 58 BAC 7 40 MAC 23 33 QAC 10 10 -8- 2. Joint Allocation and Nestina Within O-D/Segment * Published by Curry in Transportation Science (1990) -- jointly finds optimal allocation to each "O-D nest" and nested limits on each booking class within the nest approach solution to the formulated problem -- optimal -- still a discrete allocation of seats to each segment/O-D (i.e., "O-D nest") - Both approaches above can have positive revenue impacts compared to leg/booking class control provided that number of discrete O-D nests does not become large e Otherwise, negative revenue impacts can result when implemented as a control methodology 1NIMMINIM1111W 161 k1i NNINSINN hi ,11131 111111 11110 1 1j,11191M 1 Will, -9- Prices of ODF Allocations on Shadow Nestinga 3. Shadow Nestina on 3. * Described in Williamson's (1988) MIT Master's thesis ODF allocations a re ranked and nested in order of shadow price values derived fro n the optimization algorithms -- optimal -- the shadow -- on each flight leg, ODFs with highest price of an ODF allocation the amount by which total expected network revenue will increase (or decrease) if one additional seat is allocated to that ODF shadow prices values receiv e greatest availability -- ODF allocations are treated as "protection levels" for nesting purposes -10- EXAMPLE: 0 DF YAB BAB YAC BAC MAB MAC QAC QAB e 2 segments on Leg A-B (Capacity = 100) Optimal Allocation 8 12 18 7 6 23 10 16 Shadow Price 225 200 190 165 110 40 10 0 Nested Limits 100 92 80 62 55 49 26 16 Implementation into control structure possible through virtual inventorv classes defined by shadow price ranges Nesting of optimal ODF allocations on current shadow prices results in theoretically sub-optimal booking limits for different ODFs e Yet, a large number of dynamic booking simulations of this approach as a control methodology show consistent and substantial revenue improvement over leg/booking class control -ll- Network 4. 4. Network * "Bid-Price" Developed at MIT: Williamson (1990) Approaches Simpson (1989) and optimization can also produce shadow prices on the capacity constrain associated with each flight leg, or "bid prices" -- network -- Bid Price Control of seat inventories simply requires a comparison of the fare of the requested ODF itinerary and the sum of the bid prices involved in the itinerary. -- Implementation requires frequent (real- time?) updating of network bid prices to overcome absence of booking limit controls * Bid Price Approach is a sub-optimal control application of optimal network solution, which over comes negative impacts of discrete ODF allocations through "implicit nesting" of availability through bid price evaluation decision. -12- SUMMARY * OPTIMIZATION APPROACHES NETWORK -- Network optimization approaches produce optimal ODF seat allocations over a network of flights: -- require ODF demand forecasts and fares -- * partitioned ODF allocations that must be "nested" in sub-optimal ways to have positive revenue impacts generate Truly optimal solution for control of ODF itineraries over a network requires an approach that for nesting of ODFs explicitly -- accounts -- allows "desirability" -- recognizes -- overcomes "small number" problems of of ODFs to change as demand materializes dynamic nature of future booking process forecasting ODF demands -13- CONCLUSIONS 0 Network optimization methods produce an optimal solution to an assumed mathematical formulation of the O-D/segment control problem Implementation of optimal solutions in actual reservatior is control structures can lead to negative revenue impacts if done incorrectly * Nesting of optimal network ODF allocations for control purposes is a sub-optimal solution, although some nesting approaches consistently produce better revenue impacts than others * No one has formulated, let alone "solved" the dynamic, nested ODF network seat inventory control problem -14- APPLICATION OF NETWORK SOLUTIONS TO O-D SEAT INVENTORY CONTROL Elizabeth L. Williamson Flight Transportation Laboratory Massachusetts Institute of Technology Presented to MIT/Industry Cooperative Research Program May 23, 1991 Cambridge, MA -t -15- Introduction Reviewing network seat inventory control techniques and applying them to three different multi-leg examples, using real airline data: 1) 2 Leg Flight A B C B C 4 Fare Classes 3 OD Pairs 12 ODF Combinations 2) 3 Leg Flight D 4 Fare Classes 6 OD Pairs 24 ODF Combinations 3) 4 Leg Flight e- - A 4 Fare Classes 10 OD Pairs 40 ODF Combinations e- B - + C - + --. D E -16- Network Solutions Nested on Shadow Prices - Network formulation used to find seat allocations for each ODF over an entire network of flights. - Distinct allocations are nested according to the shadow price of each ODF. *Shadow Price: The amount the optimal system revenue value would change if one more seat was made available to the given, ODF. -17- Nested Deterministic by Shadow Prices 3 Leg Example Leg BC - Capacity=90 ODF ACY BCY ADY ACM BCM BDY ACB ACQ BCB ADM BCQ BDM ADB ADQ BDB BDQ Seats Allocated 2 10 3 1 22 6 4 14 12 1 15 0 0 0 0 0 Fare 519 440 582 344 315 440 262 231 223 379 197 307 302 269 Shadow Price Booking Limit 322 243 216 147 118 90 88 78 74 52 46 65 34 26 75 74 42 13 28 16 0 15 -59 -64 221 -97 -145 199 -167 0 0 0 0 0 -18- NETWORK SOL'NS NESTED ON SHADOW PRICES 2 Leg Flight 2.6 2.4 l 22 w 2 V 1.8 1.6 u 1.4 0 1.2 1 C 0.8 0.6 16- 0.4 S. 020 -02 -0.4 0.7 0.74 0.78 0 0.82 0.86 Load Factor NDSP + NPSP 0.9 0.94 0.98 -19- ON SHADOW PRICES NETWORK SOL'NS NESTED 3 Leg Flight 4 (12 Li N 3 2 U 61 2 0 S.. '.4 0. 61 U 63 S.. 61 '.4 '-S 4.' U U Li U a. 0.76 0.84 0.8 0 0.88 Load Factor NPSP + NDSP 0.92 0.96 -20- NETWORK SOL'NS NESTED ON SHADOW PRICES 4 Leg Flight 0.55 0.85 0.75 0.65 Load Factor 0 NDSP + NPSP 0.95 -21 DISTINCT DETERMINISTIC APPROACH o00 500 -L 400 300 200 100 0 0 5 10 15 20 25 30 35 DISTINCT PROBABILISTIC APPROACH 500 500 400 300 200 100 0 0 5 10 15 20 25 30 35 3 13 EMSR APPROACH 00 500 400 300 200 100 00 5 5 10 10 15 15 20 25 20 25 - -22- EXAMPLE Single Leg, 4 Fare Classes MEAN 20 15 30 25 Y M B Q STD 7 5 10 8 FARE 500 350 200 150 ALLOCATIONS DETER Y M B 20 15 30 25 Q PROB EMSR 27 19 31 23 17 20 27 36 BOOKING LIMITS Y M B Q NDSP NPSP 100 80 65 35 100 73 54 23 EMSR OPTIMAL 100 83 63 36 100 83 62 33 -23- Initial Allocations 3 Leg Example AB Leg - Capacity=75 Distinct Probabilistic Distinct Deterministic: AB AC AD Y M B Q 25 2 3 3 1 0 7 4 0 26 4 0 AB AC AD Y MB Q 28 3 2 5 10 0 0 1 0 26 0 0 -24- Difference in Allocations (Prob - Deter) Y AB AC AD l 1 MB Q 1 0 0 -4 0 -3 0 -25- Comparison of Allocations Over 15 Revisions Mean AB Y 25.2 25.1 24.8 24.0 22.8 22.0 20.4 19.3 16.9 15.6 12.3 9.2 8.6 5.9 2.6 Deter Alloc Prob Alloc 25 25 25 24 23 22 20 19 17 16 12 9 28 28 28 28 28 26 26 26 25 23 21 19 9 18 6 3 15 11 -26- Partially Nested versus Fully Nested Partially Nested (Curry): e Determine discrete allocations for each OD, based on expected revenue from nested fare classes. *Determine fare class booking limits within each OD allocation. lili UNwill. -27- Expected Revenue per O-D Pair BC 460 440 420 400 380 360 340 320 300 280 260 240 220 200 180 160 140 Seat -28- Fully Nested versus Partially Nested 3 Leg Example Leg BC - Capacity=75 ODF NDSP Allocations NDSP BL DOD-NFC BL ACY ACM ACB ACQ 2 1 4 4 75 4 3 ADY ADM ADB ADQ 3 0 0 0 BCY BCM BCB BCQ 10 22 12 11 BDY BDM BDB BDQ 6 0 0 0 38 31 15. 0 0 63 2 0 0 0 0 0 0 73 64 57 60 27 15 37 0 0 0 38 32 5 1 0 0 -29- FULLY NESTED VS. PARTIALLY NESTED 2 Leg Flight 2.6 22 2 1.8 1.6 1.4 12 1 0.8 0.6 0.4 02 0 -02 0.7 0.74 0.78 0 0.82 0.86 Load Factor DOD-NFC + NDSP 0.9 0.94 0.98 -30- FULLY NESTED VS. PARTIALLY NESTED 3 Leg Flight 3.5 3 C,, :22 VZ. E oas 1.5 0 L. .8.9 0 Load Factor DOD-NFC + NDSp -31 - FULLY NESTED VS. PARTIALLY NESTED 4 Leg Flight 2.5 0 -0.5 -1 - 1.5 0.55 0.65 0.75 Load Factor 0 NDSp + DOD-NFC 0.85 0.95 -32- Bid Price e Bid Price is a Shadow Price for the capacity constraints. - Obtained from the same network formulations. - The marginal value of the last seat of a given flight leg. - Bid Prices establish a "cutoff" value for each flight leg, on which decisions can be made whether to accept or reject a given O-D/fare class request. - For a single leg itinerary, a fare class is open for bookings if the corresponding fare is greater than the bid price, or shadow price, for the leg. *For a multi-leg itinerary, fares must be greater than the sum of the bid prices from the respective flight legs. MI -- No -33- 3 Leg Example Capacity= 7 5 B A A-B B-C C-D C 34 197 169 BC: 197 AC 231 y M B 440 Y M B Q 519 344 262 231 Q 315 223 197 D AD: 400 582 379 302 269 -34- DETERMINISTIC NETWORK METHODS 2 Leg Flight 2.6 2.4 22 Z2 1.8 1.6 1.4 12 1 0.8 0.6 0.4 0.7 0.74 0.78 0.82 0 ~ 0.86 Load Factor BID + NDSP 0.9 0.94 0.98 m -35- SOLUTIONS DETERMINISTIC 3 NETWORK Leg Flight 32 3 2.8 2.6 2.4 22 3 2 - 1.8 0 1.6 1.4 C 12 - 0.8 C u 0.6 0.4 02 a 0.76 0.84 0.8 0 0.88 Load Factor BID + NDSP 0.92 0.96 -36- DETERMINISTIC NETWORK METHODS 4 Leg Flight 2.6 2.4 22 2 1.8 1.6 1.4 12 1 0.8 0.6 0.4 0.55 0.75 0.65 0 Load Factor + BID NDSP 0.85 0.95 -37- Revenue Impa cts vs. Revisions 3 Leg Example 1 - 98% Load 3 Number of Revisions + BID NDSP 0 Factor -38- PROBABILISTIC NETWORK METHODS 2 Leg Flight -1.5 -3.5 0.7 0.74 0.82 0.78 0 0.86 Load Factor PBID + NPSP 0.9 0.94 0.98 ow fifillh -39- PROBABILISTIC NETWORK SOLUTIONS 3 Leg Flight -0.5 -1.5 - 2.A 0.76 0.84 0.8 C 0.88 Load Factor PSIC + NPSP 0.92 0.96 -40- PROBABILISTIC NETWORK METHODS 4 Leg Flight 2 1 a -1 0.55 0.65 0.75 0 Load Factor PBID + NPSP 0.85 0.95 1,1111milmmillbi -41- UPPER BOUND MEAN 36.12 9.94 18.61 34.06 ABY ABB ABM ABQ STD DEVI 14.91 13.92 16.01 - 25.96 ACTUAL DEMAND ABY ABB ABM ABQ 35 14 18 39 28 12 18 32 48 8 16 -42- UPPER BOUND COMPARISON 2 Leg Flight 0.7 0.74 0.78 0 NDSP 0.82 0.86 Load Factor 0 + BID 0.9 UPPER 0.94 0.98 __________N -43- UPPER BOUND COMPARISON 3 Leg Flight 0.76 0.8 0.84 0 NDSP 0.88 Load Factor 0 + BID 0.92 UPPER 0.96 -44- UPPER BOUND COMPARISON 4 Leg Flight 0.65 0.55 0 NDSP 0.815 0.75 Load Factor 0 + BID UPPER 0.95 U -45- Summary * Nested Deterministic on Shadow Prices outperforms Nested Probabilistic on Shadow Prices. - Given full ODF forecasts, better to use a fully nested method, such as NDSP, rather than a partially nested method. e Deterministic Bid Price approach performs well and uses a very simple control methodology, however it is important to be able to make frequent revisions using such an approach. - Using Upper Bound, the true potential from better control of seat inventories over current leg based approaches can be determined. Mhmssclwsetts Institute of Technology Flight Transportation Laboratory-g Planning and Scheduling of Tasks ma Dynamic Environment Lyman R. Hazelton 23 May 1991 LRW9I-OO1 1VMassachusetts Flight Transportation Laboratory-g Institute of Technology The Strategic Control of systems requiring planning and scheduling of activities is called Operations Management Reasoning about the future in a dynamic environment. Is Determination of the time that a state or process should be maintained. I Situation dependent objectives. I No final system State. I Often involve non-quantifiable parameters. LRH-90-04" Flight Transpotto Institute of ~u~~i-~- Adecision was made to attempt to solve the problem with an "Expert Systems" approach. However, existing Al planning methods * Were based on a back'-chained, goal seeking technology. Have been shown to be NP-hard or even Non-terminating for. conjunctive goals. Assumed a single actor, non-stochastic universe. * Had no logic or even 'representation for time, dependent activities. In summary, the automatic reasoning technology necessary to attack the problem did not exist. LRH-90-043 Massachusetts Institute of T'-.chnology Flight Transportation Laboratory-- At the time the research was initiated: I I There were NO prograpis or even algorithms for temporal database management There were NO data representations for concurrent temporally bo-anded, information Automza± plan generatior. was restricted to Single Actor Determinate Dcmains Instantaneous Actions LRH4-90-050 Mesacwt Ingtitute of Flight Transportation Laboratory Operations Management Model LRII-90-{) 1'" Technology Flight Transportation Laboratory Advice Observations A Y Temporal System Analyzer RULESYS (Logical Inference Engine) Assertions Schedules Requests Y ASchedules Time Map Manager (Temporal Database Manager) TIMEBOX SCHEDULE LRH-9(~O(' Technogoy Flight Transportation Laboratoryl Pattern-I Search Section Pattern-3 Antecedent Test Section Rule Consequent Pattern-2 Test-1 Test-2 Assertion-1 Assertion-2 LRH-9 Technolo Flight Transportation Laboratoryg Plan: PAINT LADDER Plan: PAINT CEILING Procedure: GET PAINT GET LADDER APPLY-PAINT LADDER Procedure: GET PAINT Goal: NEAR CEILING APPLY-PAINT CEIIJNG Results PAINTED LADDER Rcsults PAINTED CELMN / LRH-910- Masschuset isuen O Flight Transportation. Laboratory SHIT Pan: PAINT CEILING Han: PAINT LADDER Procedur GET PAINT Goal: NEAR CEILING APPLY-PAINT CEILING Proceduim GET PAINT GET LADDER APPLY-PAINT LADDER Results: PAINTED CEILING Results: PAINTED LADDER LRHI-91-IY Iitofd Technology Flght Transportation Laboratoryg SKH LIAction: GET PAINT Action: GET PAINT Action: GEf LADDER Actiom GET LADDER Action: CLIMB LADDER Action: APY-PAINT LADDER L Action: GEr PAINT izzirizzi] Action: GET LADDER D Action: APPLY-PAIN T CEILING [Action: CLIMB LADDER KIN L Action: APPLY-PAINT LADDER _Action: APPLY-PAIN r CEIIuNG Jom LRH-91-0)i Masschuisets Institute of Techftology Flight Transportation Laboratoryg Truth Maintenance A first attempt to extend logic into a dynamic environment. ,R:O p - r 3R p q}T r q p ~q R p T V r r q LPKH-90-006 Institute of Flight Transportation Laboratory * Inferred evolution R:p -q p .+r-p- q {R p q} Tr r (TM) q Rp r p Itieogy*gh Transportation Laboratory~± Introduce EXPLICITLY the TIME INTERVAL during which a proposition (was, is, will be) true: p (T) where r is a time interval having a starting time and an ending time. LRII-90-011 Massachugetts Institute of Technology Flight Transportation Laboratory Persistence: R p ~p 5 q q (NO TM) If it is raining (P), the roads will be wet (q). But if it stops raining CP), the roads do not instantly become dry. Wet roads persist. LRH-90-010) Masmichugetts Institute of -Techfiolo~v Flight Transportation LaboratoryI.1 Temporal Logic (continued) * Rules of Inference Modus Ponens: p +q ) p( Non-persistent q (start( ), o) Persistent Sg (T) 0 0 0 LRI*-90-021 Massachusetts Institute of -Technology Flight Transportation LaboratorylgInferred evolution -revisited R:p( 1 ) -q r (Tl T T)- ,p(rfl T ) 2 {R, p(T 1 ) q (T 2)} Tr (ri T2) {R p(rTI) q (T2)} T p (Tifl T2) ~p (v) qg(r) T r(T,n T) r(r) p (T, n T 2) LRI4-90-023 Mamschusetts Institute of atL boatr Flight Transportation Laboratoryg P The problem stems from the fact that the reasoner's BELIEF (i.e., knowledge) changes during the reasoning process. There are TWO tinie intervals involved in temporal reasoning. LKH-90-025 Massachusetts Institute ofio Flight Transportation Laboratorylg * The ACTIVITY interval, during which the proposition (was, .is, will be) true * The BELIEF interval, during which the reasoner believes a proposition about some activity, interval to be true LRH'-90-026 Institute of Tecbnolg light Transportation Laboratory Types of consequents Bounded Persistent_ Decayed Probabilistic LRH-90-042 -65I-. I~' BEFORE 12 ABUTS-BEFORE - -- -il I P (STARTS-BEFORE ENDS-DURING) I'-.-] (STARTS-BEFORE ENDS-EQUAL) LYILJIJ L3] CONTAINS F77T~1 (STARTS-EQUAL ENDS-DURING) J7~ EQUAL (STARTS-EQUAL ENDS-AFTER) 0 IS-CONTAINED-BY I_ rx'zrn I (STATS-DURING ENDS-EQUAL) L H .J1Z17.. 1, *- ~ '-U--n (STARTS-DURING ENDS-AFTER) -1- ABUTS-AFTER [~j7fl AMTER Figure 3.2: Temporal Relations 6-8 ' .' ZHj Institute of Technology Flight Transportation Laborator LRH1-904)41 Masschusetts Institute of Technlg Flight Transportation Laboratory FACT: thing attribute Future Past t t2 LRH-90-040 -68- Most Recent Belief Earliest Belief History Belief Interval - Owning Fact Cell List I terval U PS4art Time EImdA, Time U History Cell Interval S ftL Owning History Active Interval Value List Tin m1 End Time Value Cel Value Sense Assumption LRH-90-040 Massachusetts Institute of Technology Flight Transportation Laboratory-g CONTRIBUTIONS Extension to Non-monotonic Temporal Logic by introducing Belief Intervals Introduced Persistence as rule specific knowledge Designed structures to represent time dependent knowledge Implemented an efficient temporal database management program LRH-90-065 Massachusetts Institute of Technology Flight Transportation Laboratory# CONTRIBUTIONS (continued) e Implemented a Temporal System Analyzer employing Extended Temporal Logic and Persistence Created a Scheduler Program, thereby extending Domain Independent Planning to include Parallel, Time Bounded, Non-Instantaneous Actions LRH-90-06( Massachusetts ofi"' T&e-hnoog Flight Transportation Laboratory- g' Novel ideas and methods developed for this system include *0 A highly compact representation for the description of descrete time dependent processes. * An efficient time based logical inferrence system. Deeper understanding of human cognitive and communication processes involved in Command and Control Systems. * A replacement of "Truth Maintenance" by "History Maintenance", and a better understanding of default versus dynamic logic. LRH-90-044 Massachusetts Institute of TectInology Flight Transportation Laboratory' 0 LRH-90-068 Concentration in U.S. Air Transportation: An Analysis of Origin-Destination Markets since Deregulation Jan Van Acker Flight Transportation Laboratory May 23, 1991 Agenda I. Thesis Objective and Methodology II. Analysis of Top 100 Markets III. Analysis of Dominated City Markets IV. Conclusions I. Thesis Objective e Study effects of deregulation on concentration - Focus on Origin-Destination City-Pair Markets Focus on Concentration in O-D City-Pairs e e Other studies found: - Fares are positively related to concentration - Concentration levels have decreased on average Our study looked at: - Top 100 domestic O-D markets - Markets out of dominated cities Measurement of Concentration e Concentration indices used: - Hirschman-Herfindahl Index (HHI) - - e 2-Firm Concentration Ratio (C2) Number of Competitors with >5% Market Share (Number of Effective Competitors) Market share is measured in terms of local passengers transported in market II. Changes in Concentration in Top 100 Markets e Markets ranked 1-100 in terms of local passengers transported in 1989 e Cumulative number of passengers was 31 % of U.S. domestic total in 1989 e Years studied: 1979, 1981, 1983, 1985, 1987, 1989 e With focus on 1979, 1985, 1989 Average Number of Effective Competitors was One more in 1989 than in 1979 Year Average Number of Effective Competitors 1979 1981 1983 1985 1987 1989 2.7 3.3 3.5 3.8 3.6 3.7 4.51- 2.5EF 1.5 F 1979 1981 1985 1983 Year 1987 1989 56 Markets Were Served by Four or More Effective Competitors in '89, as Compared to only 16 in '79 # Carriers With >5% MS 1979 1985 1989 1 2 3 4 8 38 38 11 5to6 5 7to8 0 0 17 30 24 28 1 1 19 24 29 24 3 501r 40 30 gd 10 0 1979+ 1985* 1989 # of Carriers with >5% Market Share 62% of the Passengers Flew in Markets Served by 4 or More Effective Competitors in '89 - - only 18% in '79 # Carriers With >5% MS 1979 1985 1989 1 2 3 4 5 to 6 7 to 8 5.5% 30.9% 45.8% 11.3% 6.5% 0.0% 0.0% 12.6% 29.1% 26.3% 30.7% 1.2% 0.7% 16.5% 20.5% 28.9% 30.2% 3.3% 40% 30% 10% 0 6 4 2 a 1979+ 1985* 1989 # of Carriers with >5% Market Share 8 10 Average HHI Was Lower in 1989 than in 1979 Year Average HIHI 1979 1981 1983 1985 1987 1989 4917 4077 3913 3361 3705 3586 5.515F U 'U 4.5 F- 0 3.5 F 2.5 1979 1981 1983 Year 1985 1987 1989 The Majority of the Markets Experienced a Decrease in HHI from 1979 to 1989 Change in HHI -8000 to -6000 -6000 to -4000 -4000 to -2000 -2000 to O 0 to 2000 2000 to 4000 4000 to 6000 Total Decreased Total Increased Average Change -7 -? o -5 -3 3 T-1 1 3 (Thousands) 1979-1989 1979-1986 1985-1989 Change in HHI The Non-Hub Markets Were Served on Average by a Greater Number of Effective Competitors in '89 than the Hub Markets Year Hub Markets Non-Hub Markets 1979 1981 1983 1985 1987 1989 2.7 3.3 3.7 3.9 3.5 3.5 2.6 3.4 3.4 3.7 3.7 4.0 4.5 3.51F 31- 1.5 F I I 1979 1981 1983 1985 1987 0 Hub Markets Non-Hub Markets + 1989 Concentration Decreased from '79 to '89 in All but One of the Non-Hub Markets '85-'89 '79-'89 '79-'85 -4to-3 -2to-1 0 1 to 2 3to4 5to6 0 1 8 33 7 0 0 5 11 24 9 0 0 11 19 17 2 0 Total Decreased Total Increased 1 40 5 33 11 19 Average Change 1.10 0.33 1.43 Change in # Carriers With >5% MS But Was Higher in '89 than in '79 in 30% of the Hub Markets Change in # Carriers With >5% MS '79-'89 '79-'85 '85-'89 1 11 9 4 1 0 7 8 28 8 0 1 26 16 8 0 0 Total Decreased Total Increased 12 30 7 36 27 8 Average Change 1.12 -0.41 0.71 -4to-3 -2 to -1 0 1 to 2 3to4 5to6 25 The Top 10 Markets Were on Average Less Concentrated than the Top 50 and Top 100 Markets Year Top 100 Markets Top 10 Markets Top 50 Markets 1979 1981 1983 1985 1987 1989 86.6% 79.2% 78.4% 73.3% 74.7% 73.6% 79.5% 74.8% 75.0% 70.6% 72.2% 66.2% 83.9% 77.7% 77.0% 71.8% 74.3% 73.3% 100% 95%90%85%80%75%70%65%60% 1979 1981 1983 1985 1987 o Top 100+Top 10*Top 50 Year 1989 Conclusions of Top 100 Markets Analysis e Average concentration was lower in '89 than in '79 e Concentration was lower in 70% of the markets e Non-hub markets were better off on average in 1989 than hub markets e Top ten markets were less concentrated on average than top 100 markets III. Changes in Concentration in Top Ten Markets out of Dominated Cities e Cities at which 60% of total passenger enplanements in 1985 were carried by one airline, or 85% by two: Atlanta Charlotte Cincinnati Dayton Denver e Detroit Greensboro Memphis Minneapolis Nashville Pittsburgh Raleigh/Durham St. Louis Salt Lake City Syracuse Markets ranked 1-10 in terms of local passengers transported in 1989 out of each of the cities Changes in Concentration in Top Ten Markets out of Dominated Cities e Years studied: 1979, 1981, 1983, 1985, 1987, 1989 e With focus on 1979, 1985, 1989 Average Number of Effective Competitors in 150 Markets Peaked in '85, but Was Still Higher in '89 than in '79 Year Dominated Airport Markets 1979 1981 1983 1985 1987 1989 2.2 2.8 2.8 3.1 2.9 2.5 4.5 4 3.51F 3F 2.51F 0", 1979 1981 1983 Year 1985 1987 1989 Average Number of Effective Competitors for each of the Dominated Cities Top 10 Atlanta Markets Top 10 Charlotte Markets 3. 4.5 4.0- 4.0- 3.5- 3.5- 3.0 3.0 ............. ........... .- - v 2.52.0- 2.52.01.5- 1.0' 5.0 1979 'op 1981 1983 1985 Year 1987 1989 1979 10 Cincinnati Markets 1981 1983 1985 * Year 1987 1989 Top 10 Dayton Markets 4.5- 4.5- 4.0 4.0- 3.5 - 3.5 3.0 3.0- 2.5 2.5- e--"............................... 2.0 t2.0- 1979 1981 1983 1985 Year 1987 1989 1979 1981 1983 1985 Year 1987 1989 Average Number of Effective Competitors for each of the Dominated Cities Top 10 Denver Markets Top 10 Detroit Markets 4.5 4.0- 4. 3.5- 3. 3.0 3.0 2.5 2.52.0- 2.05 1979 1981 1983 1985 Year 1987 1989 0 1.4 1979 Top 10 Greensboro Markets 5.Oi- 1981 1983 1985 Year 1987 1989 Top 10 Memphis Markets Or4. 5 4.0 3.0 3.5 [ 0 2.5- . ....... .. ... .......... 2.5 2.01. 1979 1981 1983 1985 Year 1987 19899 1979 1981 1983 1985 Year 1987 1989 Average Number of Effective Competitors for each of the Dominated Cities Top 10 Minneapolis Markets Top 10 Nashville Markets 5.0 4.51 4.0 1 3.01. 2.546/ 1979 1981 1983 1985 1987 1989 1979 1981 Year Top 10 Pittsburgh Markets 5.%#1 To~ To 4.5 4.5 - 4.0 4.0- 3.5 33.5- 3.0 3.0- 2.5 n 2 10 1983 Year 1985 1987 1989 Raleigh-Durham Markets 2.5- __z 2.0 2.0- 1.5 1979 1981 1983 1985 Year 1987 1989 1979 1981 1983 Year 1985 1987 1989 Average Number of Effective Competitors for each of the Dominated Cities Top 10 Salt Lake City Markets Top 10 St. Louis Markets 5.0 4.5- 4.5 4.0 4.0- 3.5- 3.5 - 3.0- 3.01- 2.5- 2.5 2.0- 2.0 1.5 1.5 1n - . 1979 1981 1983 1985 Year 1987 1989 1979 1981 1983 1985 Year Top 10 Syracuse Markets 1979 1981 1983 1985 Year 1987 1989 1987 1989 Changes in Concentration in the Top Ten Atlanta Markets O-D City-Pair Markets Atlanta Atlanta Atlanta Atlanta Atlanta Atlanta Atlanta Atlanta Atlanta Atlanta Boston Chicago Dallas/Fort Worth Los Angeles Miami New York Orlando Philadelphia Tampa Washington HHI 1989 '79-'89 '79-'85 '85-'89 5446 2949 5932 5089 3737 3913 5608 4097 6002 4701 455 -2538 -350 -219 -1004 -935 482 -995 955 -193 -547 -2031 -1968 -656 -885 -1294 -880 -410 -637 -721 1002 -508 1618 437 -119 359 1362 -585 1593 528 7 3 10 0 3 7 -434 -1003 569 Total Decreased Total Increased Average Change in HHI 4747 Concentration Levels Decreased Substantially in Most of the Top Ten Syracuse Markets O-D City-Pair Markets Syracuse Syracuse Syracuse Syracuse Syracuse Syracuse Syracuse Syracuse Syracuse Syracuse Atlanta Boston Chicago Detroit Los Angeles New York Orlando Philadelphia Tampa Washington HHI 1989 '79-'89 '79-'85 '85-'89 4418 9045 4119 8942 1585 5820 3047 9741 2695 8289 -5552 175 -5473 3967 -5234 653 -6014 790 -5875 -1406 -4065 -4091 -5889 -740 -4924 -1756 -5507 493 -5178 -4879 -1487 4266 417 4707 -310 2409 -507 296 -697 3473 6 4 9 1 4 6 -2397 -3653 1257 Total Decreased Total Increased Average Change in HHI 5770 Concentration Increased in all Top Ten St. Louis Markets after the TWA-Ozark Merger O-D City-Pair Markets St. Louis St. Louis St. Louis St. Louis St. Louis St. Louis St. Louis St. Louis St. Louis St. Louis Chicago Dallas/Fort Worth Denver Detroit Houston Los Angeles New York Phoenix San Francisco Washington HHI 1989 '79-'89 '79-'85 '85-'89 3347 4528 4671 3306 3561 5486 8860 4780 6567 8252 -2161 -506 -238 -2180 -916 -48 2650 -320 155 302 -3019 -2037 -2333 -2200 -1626 -654 -1356 -1360 -547 -3798 858 1530 2095 20 710 606 4006 1040 701 4100 7 3 10 0 10 -326 -1893 1567 Total Decreased Total Increased Average Change in HHI 5336 Conclusions of Dominated City Markets Analysis e Single trend of hub development led to decreased concentration through '85 at most of the cities, but to increases from '85 on e Two-hub markets were less concentrated than one-hub markets in 1989 * Average concentration across the 150 markets was slightly lower in '89 than in '79 IV. Conclusions e Concentration was lower in top 100 markets, both on average and in most of the markets e Concentration in non-hub markets decreased throughout period '79-'89 because of development of hub-and-spoke networks " These networks led to increases in concentration in most hub markets after 1985 CD Conclusions e Single trend of hub development led to decreases in concentration through '85 at most of the dominated cities, but to increases from '85 on e Concentration was on average slightly lower in the 150 markets out of dominated cities in 1979 than in 1989, and was lower in half of the markets -102- Pricing in the Airline Industry Current Practice and Future Research Theodore C. Botimer MIT Flight Transportation Laboratory Presentation to Cooperative Research Program May 23, 1991 Ihi, I1, 1111 10111NIM1111 kilmhIIIW1,W -103- Presentation Outline Overview * Nature of Airline Competition * Fare Product Differentiation - Seat Inventor y Management * Pricing Strategies - Role of the Pricing Analyst - "The Ultimate Pricing Model" * Theoretical Issues for Investigation Case Study Analysis * Case Study Overview * Case Study Objectives * O/D Market Choice - ATL - BOS Market - ATL - STL Market - Conclusions -104- Nature of Airline Competition - Hub and spoke route structures prevail in the industry allowing almost every major carrier to serve any O/D market - Most competition on non-price level - Dollar value of nonstop service is unclear - Must consider strength of competitive position in e ach O/D market separately - Characterize c ompetitic n in all markets: - major players level of service offered - number of flights p r day offered - nonstop vs. nonsto p competition e - Anticipate r espor se to price changes: * who ai oe the competitors? * do the competitors offer comparable service in the market? - how have competitors reacted to fare chan ges in the past? - what response will be given to hostile reactions by competitors? MOMME1111" ,1,1,, illdilill III, -105- Fare Product Differentiation - Airlines seek to segment demand by off ering differentiated fare products in different fare classes - Delta offers tickets in 10 fare classes: 1) F - full fare first class 2) Y - full fare coach class 3) B - reserved for military/ convention/negotiated fares 4) M - highest discount coach fare 5) H - discount coach class fare 6) Q - discount coach class fare 7) K - reserved for competitive filings 8) L - reserved for competitive filings 9) A - first class free tickets 10) W - coach class free tickets - Differentiation occurs within fare classes * peak vs. off-peak fares - weekday vs. weekend fares -106- Fare Product Differentiation (con't) - Fare restrictions or "fences" used to control which type of consun er is able to purchase which type of ticket - fare restriction s include: advanced purchase requirements Saturday night stayover blackout periods flight validity resi:rictions (good for travel between...) ticket purchase r estrictions (purchase ticket s by...) availability limits for discount fares military discount fares senior citizen discount fares Common - ..... WANNME, -107- Seat Inventory Management (IM) - Pricing sets O/D prices and restrictions - IM decisions made with fixed prices and restrictions - IM seeks to maximize revenue given fixed prices and restrictions - IM controls price/seat quantity decisions - protect full far'e seats - limit discount fares - strictly limit deep discount fare seat availability - Matching stances require booking limits - strictly limited availability on competitive far e filings -108- Pricing Strategies and Their Effects - Matching a fare - retain market share possible drop in yield remain listed on Page 1 of CRS often done to remain competitive viewed as price taker in the market - Not matching a fare - possible loss of market share maintain yield may lose competitiveness loss of goodwill - Partially matching a fare e - attempt to retain market share reduce non-matching yield loss market factors influence strategy will be non-competitive at peak accept that competitor offers low fare on all flights WWSMN WIININ 1114 , ,, IIN i, -109- Role of the Pricing Analyst - Analysts do not look at operating costs - Consider strength of competitive position in each O/D market - Add routing restrictions to discount fares - Pricing analysts should be familiar with own market and relevant hub: - traffic flows - flight load factors - Be aware of fare differential effects - high differentials not seen on CRS - business travelers susceptible to higher differentials - not all fares registered in ATP listings are available in reality - Must monitor the number of bookings to determine the effect on yield of changes -110- Ultimate Pricing Model - Inputs: - published daily fare changes - system-wid[e flight schedule - price level (by O/D market & flight) - Outputs: - Su " - ested mach gb strategy ~ in g -- partial matching -- not matching - Projected impact on - Projected impact on - Management reports -suggested match market share revenue telling: ing decision -- implemented matching decision for matching decision - Ability to run simulations - Ability to do what-if scenarios -- reasons NwA 1011ME111 1, -111- Theoretical Issues for Investigation - Joint seat/price optimization problem - Optimal differential pricing strategies - Model development for pricing strategies - matching - not matching - other pricing strategies of price changes - Impacts - Measurements of price elasticity - Explore impacts of pricing strategies on: - profitability load factor yield customer satisfaction -112- Case Study Overview - Close look at 10 O/D markets - Representative cross section of markets flown by Delta Airlines - Quarterly analysis - Examine quarterly data 1986:1 - Give consideration to: 0 blI p 4 s eF6L& - 1990:2 aresa - competitive responses - major price level changes - Use information from several data bases: - PIPPS (Historical ATP data) - DOT O/D traffic stats (10% sample) - Official Airline Guide - Preliminary analysis on two markets: - ATL - BOS - ATL - STL , I, IM111 , , ''111,111MI IN I, -113- Case Study Objectives - Initial look at revenue management from pricing perspective - Develop market by market case studies - Present a market overv iew - Characterize pricing practices - Analyze competitive environment - Uncover competitive characteristics - Highlight major market events the quality and level of detail of the ava ilable data sources - Analyze - Relate - Develop market strength to fare level - between carriers - over time a measure of the sensitivity of travelers to changes in fare level - Determine selling fares during the period - Use available data to determine the effects of pricing decisions - Discuss future directions for research -114- O/D Market Choice Lenath of Haul - Short Haul (<1000 miles) Medium F aul (1000-2000 miles) - Long Haul (>2000 miles) - Nature of Competition of Competition - Delta offers non-stop service Only competitors offer non-stop service - No one offers non-stop service Markets Chosen 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) ATL-BOS ATL-SEA ATL-STL BOS-PHX CLT-MSP DFW-PHL JAN-SDF MSP-SAN MSY-PWM SAV-SAN -115- ATL - BOS Market Characteristics 1986:1 - 1990:2 - Two non-stop carriers during the period -- Delta -- Eastern - Non-stop carriers flew 93% of all pax - Frequency of approximately 12 daily non-stops each way - Total traffic leve l of 925 1passengers per day in both directions - Carriers with ATL hub -- Delta -- Eastern - Eastern Airlines strike in 1989:2 -116- ATL-BOS Passengers 1986:1 - 1990:2 11 10 - 9 I 8 8 8 7:3I I I I I I I I I I89: 90: I 89:1 89:3 90:1 88:3 86:3 87:1 87:3 88:1 86:1 90:2 88:2 88:4 89:2 89:4 87:2 87:4 86:2 86:4 YearQuarter ,, 111h B11141111111 kllilfillllll -117- ATL-BOS Average Fare 1986:1 - 1990:2 220 210 200 190 180 170 160 150 1140 I 130 86:1 863 86:2 87:1 86:4 87:3 87:2 88:1 87:4 88:3 88:2 Year.Quarter 89:1 88:4 I 89:3 89:2 t I 90:1 89:4 90:2 MVLBOS Morke~ /1A Shore 8OL& 60 If) nfl 24 20v 10 -A 86:1 ~ 86:2 86:3 I I 86:4 87:1 -- 87:2 87:3 87:4 88:1 882 88:3 88:4 89:1 YearQuarter -,- Delta - - Eastem 89:2 89:3 89:4 90:1 902 /YL~HS verae rare 1986:1 260 -~---- m- - - 1990:2 _ -- - -- _- - - _ _ 240 20 140 120 86:1 86:2 86-3 86:4 87:1 87:2 Delta 87:3 87:4 88:1 88-2 88:3 YearQuarter Eastem M84 89:1 89:2 89: 894 9- 19: ATL - BOS Summary Table Passengers Delta 88:2 Delta 89:2 % Change Eastern 88:2 Eastern 89:2 % Change OA 88:2 OA 89:2 % Change Market 88:2 Market 89:2 % Change 6118 7676 25.47 3508 9 -99.74 615 1387 125.53 10241 9072 -11.41 Revenues 1152456 1584791 37.51 599469 1674 -99.72 88075 214329 143.35 1840000 1800794 -2.13 Average Coupon Passengers Market Revenue Yield Fare Mileage Per Day Share Share Per CPM 188 206 9.60 171 186 8.84 143 155 7.90 180 199 10.48 5872057 7372559 25.55 3348572 8514 -99.75 653125 1429440 118.86 9873754 8810513 -10.77 67.2 84.4 25.47 38.5 0.1 -99.74 6.8 15.2 125.53 112.5 99.7 -11.41 59.74 74.95 25.47 34.25 0.09 -99.74 6.01 13.54 125.53 100.00 100.00 0.00 62.63 88.01 40.51 32.58 0.09 -99.71 4.79 11.90 148.65 100.00 100.00 0.00 19.63 21.50 9.53 17.90 19.66 9.83 13.49 14.99 11.19 18.64 20.44 9.68 W ATL - STL Market Characteristics 1986:1 - 1990:2 - Four non-stop carriers during the period -- Delta -- Eastern -- Ozark -- TWA - Non-stop carriers flew over 90% of all pax - Frequency o f approximately 15 daily non-stops ea ch way - Total traffic level of under 450 passengers per day in both directions - - Carriers with ATL hub -- Delta -- Eastern Carriers with STL hub -- Ozark -- TWA - Eastern Airlines strike in 1989:2 - Ozark - TWA merger in 1987 -122- ATL - STL Passengers 1986:1 - 1990:2 5000 4500 8 4000 3500 30 00 ' ' f I I I- -1' l I I I I 86:1 86:3 87:1 87:3 88:1 88:3 89:1 89:3 90:1 86:2 86:4 87:2 87:4 88:2 88:4 89:2 89:4 90:2 Year.Quarter .. Ww m IN -123- ATL - STL Average Fare 1986:1 - 1990:2 180 170 160 150 140 130 120 110 100' ' ' ' 90:1 89:3 89:1 88:3 88:1 87:3 87:1 86:3 86:1 90:2 89:4 89:2 88:4 88:2 87:4 87:2 86:4 86:2 YearQuarter ATL - SIL- Market Share 1986:1 - 1990:2 50r Q- L-) 30 \ 20 :x 10 ~ 86:1 86:2 88:3 86:4 87:1 87:2 DL 87:3 87:4 88:1 88:2 88:3 88:4 Year:Quarter 2 1# -0Z 89:1 89:2 89:3 89:4 90:1 90:2 ATL - STL Average Fare 1986:1 -1990:2 IOU 170 r 160 - CDe [~' 4> en 14O 130 120 110 100 90 L .. 86:1 . i 86:2 86:3 86:4 --. I 87:1 - I 87:2 - -L 87:3 87:4 88:1 ._ - 882 _I- 88:3 -- -1 .. L0 lw ... 88:4 89:1 Year.Quarter Di El- -. 0-1 .. .. - . - - -L 89:2 89:3 89-4 90:1 _- J - 90:2 ATL - STL Summary Table Passengers Delta 88:2 Delta 89:2 % Change Eastern 88:2 Eastern 89:2 % Change TWA 88:2 TWA 89:2 % Change OA 88:2 OA 89:2 % Change Market 88:2 Market 89:2 % Change 1799 1882 4.61 1049 1 -99.90 1155.00 1631.00 41.21 67 57 -14.93 4070 3571 -12.26 Revenues 259057 320589 23.75 149500 94 -99.94 172264.00 261282.00 51.68 6003 6560 9.28 586824 588525 0.29 Average Coupon Passengers Market Revenue Yield Fare Mileage Per Day Share Share Per CPM 144 170 18.29 143 94 -34.04 149 160 7.41 90 115 28.45 144 165 14.30 881790 919694 4.30 510126 484 -99.91 560872.00 793400.00 41.46 48980 38358 -21.69 2001768 1751936 -12.48 19.8 20.7 4.61 11.5 0.0 -99.90 12.7 17.9 41.21 0.7 0.6 -14.93 44.7 39.2 -12.26 44.20 46.24 4.61 25.77 0.02 -99.90 28.38 40.07 41.21 1.65 1.40 -14.93 100.00 100.00 0.00 44.15 54.47 23.39 25.48 0.02 -99.94 29.36 44.52 51.68 1.02 1.11 8.96 100.00 100.00 0.00 29.38 34.86 18.65 29.31 19.42 -33.73 30.71 32.93 7.22 12.26 17.10 39.54 29.32 33.59 14.59 W.W w iili, -127- Conclusions - Carrier strength varies by O/D market - Delta holds a stronger position in ATL - BOS than in ATL - STL - Delta fare levels may have been too high in ATL - STL during the strike given its competitive position -128- Future Directions - Quantify consumer price sensitivity and market share chan ges - Determine relationships between market strength arnd fare levels - Develop a model to characterize competitive structure of markets -129- CHANGES IN O-D PASSENGER TRAFFIC FLOWS NEWARK AIRPORT Chung Y. Mak and Professor Peter P. Belobaba MIT Flight Transportation Laboratory_. MIT / Industry Cooperative Research Program Annual Meeting May 24,1991 -130- BACKGROUND : PREVIOUS ANALYSIS * Removal of PeoplExpress from the New York (EWR) market has had the most significant impact on traffic flows. * Domestic connecting passengers have dropped in both absolute and percentage terms at all three airports, suggesting a shift by carriers away from New York airports as domestic hubs. Newark Airport (EWR) * Stable departure levels since PE withdrawal, but fewer seats and reduced aircraft sizes. * Major drop in on-board passengers after 1986-3; downward trend continues through 1989-3 for virtually all carriers. * Local originating passengers cut by half when PE failed; levels have barely returned to pre-1984 levels. - Domestic connecting passengers were similarly affected by PE withdrawal from EWR. .- u.y . .... Ili Ii -131- Total Seats Departed for Majors Al New York City Areo Akports 12.21211.811.611.411.211 10.810.610.410.2109.8 1 9.6-1 1084 2 1 3 1 1 1 4 1085 2 1 3 4 1086 2 3 Ouarter Figure 1.2 4 1087 2 3 4 1088 2 3 4 1089 -132- Total Seats Departed for Majors (T9) Newcrk 1I*wnflon 6 5.8 5.65.45.25 4.84.64.44.24 3.6 1 2 3 1984 4 1 2 3 1985 4 1 2 3 1986 Ouartr Figure 1.8 4 1 2 3 1987 4 1 2 3 1988 4 1 -133- Total Seats Departing Nwark Interationa 3.2- 3 2.82.6 2.4 2.221.8 S=* 1.6 1.4 1.2 0.80.60.40.20 1 2 3 1984 4 1 2 3 1985 4 1 2 3 1986 Quarter O co Figure 1.9 4 + 1 PE 2 3 1987 4 1 2 3 1988 41 -1-1-- "imbMWA& -134- Total Onboard Pax for Majors Newark International 43.93.8 3.73.63.53.4- 3.3 X ac M= 03. 3.23.1 - 33- 2.92.82.7 2.2.5 2.42.3 1 2 3 1984 4 1 2 3 1985 4 1 2 1986 Quarter 3 Figure 2.3 4 1 2 3 1987 4 1 23 1988 4 1 -135- Pcx Total Onboard Nq..ar Intewngtnaw 2.1 1.9 25 1.4 - 1.7 - A' \ // ./ / / x L ~ 0.5 - 0.7 - 0.6 0.7 ~ 0.3 0.3 r.3 2 4. 2 ! - Figure 2.6 .6i 3 i 4.P z~ I4;~i 2 4. .3 4 Total Cninecting Pax (EWR) (Ten Percent Sample) 10 - 81- 1Q84 3 1Q85 3 186 3 Quarter Figure 3.5 1Q87 3 1Q8 3 1Q89 , ,, IIIIIMI IJwmIIwIiIm -137- Total Connecting Pax (EWR) 5 (Ten Percent Saample) 4.5 4 3.5 1= 00 1.5 1 0.5 0 Quarter 0 CO + Figure 3.8 PE MImIMi , -138- NEWARK AIRPORT TRAFFIC FLOW ANALYSIS (Phase 2) ANALYSIS ( hase 2) OBJECTIVE * Identify and evaluate changes in O-D passenger traffic flow patterns through Newark (EWR) and alternative hub routings. Determine shifts in connecting traffic away from EWR in O-D markets previously served by PeoplExpress. HISTORICAL DATA * Ten percent ticket coupon sample provides passenger itinerary information by quarter from 1985 to 1989. e Database Products Inc. "OD Plus" database used to extract data. - Official Airline Guide (OAG), schedule data for each of the periods. INN IIII -139- SSENGER R TRAFFIC FLOW ANALYSIS PA P ARR F N G F TRAFFIC FLOW DEMAND AND SUPPLY MEASURES - Ten percent O-D passengers travelled between each selected city pair by carrier. . Scheduled service in each city pair by carrier. AIR CARRIERS - "Major" U.S. carriers offering service to domestic destinations, defined to include smaller airlines with large market presences (e.g. Midway). -140- ANALYSIS METHODOLOGY * Obtained top 500 US Domestic O-D markets in terms of passenger traffic for 1989. * Selected markets served by PeoplExpress in 1986. - Discarded all city pairs with New York as an Origin/Destination, leaving 50 sample markets. * Used O-D Plus to obtain passenger traffic data for 3rd quarter 1986 for all major carriers serving these city pairs. * Selected O-D pairs based on market share and passenger information for detailed analysis: - markets with greater than 5% market share by PeoplExpress in 1986 or; - markets with more than 20 passengers carried by PeoplExpress per day. * A total of 20 markets were chosen based on these criteria. * Used O-D Plus again to obtain detailed passenger traffic information by individual market and carrier from 1985-3 to 1989-3. -141- 20 SELECTED O-D MARKETS CHI-BDL ORL-CMH PIT-HOU PIT-LAX WAS-MIA WAS-BUF WAS-DEN WAS-PVD BOS-CHI BOS-DFW BOS-DET BOS-FMY BOS-HOU BOS-LAX BOS-ORL BOS-PIT BOS-SFO BOS-WAS BOS-DEN BWI-DEN 1986-3 PE Share 1986-3 3.30% 2.25% 4.95% 4.91% 5.74% 9.36% 8.73% 4.99% 6.02% 4.18% 2.36% 5.16% 5.41% 1.14% 1.80% 16.45% 1.86% 3.44% 12.10% 6.45% 5440 280 3240 1810 7950 3500 7500 1670 11960 2960 2120 570 2600 1920 2110 9490 2530 9220 9320 1930 Pax/Quarter -142- FINDINGS Aggregate : 20 O-D Markets * Total traffic in selected O-D pairs decreased slightly since withdrawal of PeoplExpress in 1986-3. - aggregate traffic decreased by 5.94% from 1986-3 to 1989-3. * However, proportion of this traffic connecting through EWR dropped from 4.84% to 0.71% during the same period. - In 1985, PeoplExpress carried 8% of total traffic in these markets. - By 1989, Continental carried a total of 10% of traffic in these markets. e However, only 1% was carried by CO via EWR. ,.iiIIMEMII Nl DIAIN., -143- Total Passengers for 20 Selected Market 10% Sample 200 150 c 100 50- 0 1985 1987 Year 1986 _-_ Total * EWR 1988 1989 -144- Total Market Share 20 Selected Markets 15% 10% 0% 1985 -- _ CO - EWR O 1986 1987 Year CO - OTHER a PE A 1988 E3 1989 TOTAL - EWR -145- Disaggregate Market Analvsis Examples of market share changes 1985 to 1989 follow, showing PE, CO and the two competing carriers with the greatest increase in market share: - "CO - Other" refers to Continental traffic routed primarily through other CO hubs. -146- Market Share Comparison BOS - PIT 100.00% 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 1985 1986 _ CO - EWR C_ PA 1987 Year * CO - OTHER 1988 A PE + US BOS-PIT: PE had 28% market share in 1985, virtually all of which was taken over by USAir (non-stop service). - CO never recaptured significant market share. 1989 1161 ,1,, 11, ,11,11M INI -147- Market Share Comparison BOS - FMY 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% CO - OTHER CO - EWR o .* AA 1988 1987 Year 1986 1985 I PE DL BOS-FMY : PE had peak market share of 45% in 1985, of which CO now carries only 9% via EWR. - AA market share grew from 0 to 24% (CNX via RDU). . DL also took over market share (via ATL and CVG). 1989 -148- Market Share Comparison BOS - DEN 50.00% ... .. .... - 40.00% 30.00% 20.00% - 10.00% >0 0.00% 1985 1986 1988 1987 1989 Year _ CO - EWR -3 ML CO - OTHER + I PE UA BOS-DEN: CO has captured most of PE's 12% market share, but on non-stop service. UA also shows market share growth (non-stop service). mms =Ilml Ni -149- Market Share Comparison BOS - WAS 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% CO - OTHER CO - EWR 7,T A 1988 1987 Year 1986 1985 -- ;j PE US BOS-WAS: - PE had 10% market share in 1985. CO did not capture any of this traffic (via EWR), except in 1987 when CO offered non-stop service to IAD. Greatest MS growth by US and UA (both non-stop services). 1989 -150- Market Share Comparison CHI - BDL 80.00% 70.00% 60.00% .. ........... 50.00% dFW 40.00% 30.00% 20.00% 10.00% 0.00% 1985 1986 1987 1988 Year CO - EWR __NW CO - OTHER & PE g UA CHI-BDL: PE's 9% market share in 1985 was captured by CO via EWR until 1988, when UA increased non-stop service. 1989 WIMN 111M 'I'Allul L -151- Market Share Comparison PIT -LAX 80.00% 70.00% 60.00% 50.00% 40.00% - 30.00% 20.00% 10.00% 2 - 0.00% _ CO - EWR 1988 1987 Year 1986 1985 - CO - OTHER & PE 1_ ML PIT-LAX. . PE carried up to 5% of market share in 1986 via EWR. - CO increased its MS from 0 to 16% in 1987, but not via EWR (i.e. via IAH, DEN, CLE). 1989 -152- Market Share Comparison WAS - BUF 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 1985 1986 1987 1988 Year CO - OTHER -- CO - EWR _3 UA -- 2 PE US WAS-BUF: - PE carried 37% of market share in 1985, only 7% of which was captured by CO via EWR in 1987. - Biggest market share gains went to UA (non-stop to IAD) and US (non-stop to BWI/DCA). 1989 jLI III -153- SUMMARY OF FINDINGS O-D routings with PeoplExpress in 1986 were almost exclusively through EWR, and PE had an average of 8% MS in 20 selected markets. After withdrawal of PeoplExpress from EWR: - CO became an effective competitor in many of these markets, but traffic was split between EWR, CLE, DEN, and IAH hubs. - Growth of alternative and new hubs operated by other carriers further reduced attractiveness of EWR connections. -154- BOS CLE DEN PITEW LAX IAH O-D Routings After PeoplExpress 1989 (via Continental Airlines) -155- BOS DTW ORD PIT EWR IAD LAX RDU DFW IAH O-D Routings After PeoplExpress 1989 (Other Carriers) MCO 1114 -k- "i , - V~ j4 -156- CONCLUSIONS e Withdrawal of PeoplExpress has had significant negative impact on connecting traffic levels at Newark. - Continental took over from PeoplExpress, and Newark (EWR) became one of the 4 hubs operated by Continental with CLE, DEN, IAH. - CO now serves many O-D pairs through it alternative hubs, providing a bigger choice of departures and more direct routings. - CO did not replace PE as a competitor, its replaced PE as the hub operator of EWR. - Development of existing and new hubs by other carriers captured additional EWR market share. *IIIhI.,,,ihMEIhImUI, -157- Airline Seat Inventory Control For Group Passenger Demand Presented by Peter Belobaba Tom Svrcek May 1991 -158- Individual Passenger Seat Inventory Control Assumes Demand For Each Individual Fare Class Is Independent And Normally Distributed. B Class Y Class Q Class M Class Definition Expected Marginal Revenue (EMR) Of An Additional Seat Allocated To A Particular Fare Class Is EMR(i) = Fare Class Revenue * Probability of Selling Seat i. 15 15 ) = .05 P(X > 0) = .999 P( X > 25 EMR(1) = $500 EMR(26) = $25 -159- Individual Passenger Seat Inventory Control Example : Setup Total Fare Classes: Aircraft Capacity : 4 100 Demand Mean Demand Stdev Average Revenue Y 14 5 380 B 12 6 320 M 35 10 270 42 12 220 Fare Class Q -160- Individual Passenger Seat Inventory Control Results Seat No. Highest EMR 312.78 309.39 304.77 298.65 290.79 280.99 269.16 255.22 239.24 221.34 201.81 181.12 160.00 138.88 118.19 98.66 269.90 269.87 269.82 269.76 269.68 269.57 269.43 269.25 269.00 268.69 268.29 267.77 267.10 266.26 265.21 263.91 219.92 219.90 219.87 219.83 219.78 219.73 219.65 219.56 219.44 219.30 219.19 218.89 218.61 218.26 217.83 217.32 378.94 378.16 376.86 374.74 371.43 366.43 359.24 349.31 336.20 319.63 312.78 309.39 304.77 299.47 298.65 290.79 'Fare Class Allocations Y 16 B 13 M 34 Q 37 158.48 156.37 152.17 -161- Group Passenger Demand Why Is Group Demand Different From Individual Passenger Demand ? - Group Demand Is Realized Many Months In Advance Examples : Rose Bowl, Mardi Gras ... - Groups Negotiate For A Lower Fare (Bulk Pricing) - Unused Bookings Are Absent From Seat Inventory For Months, Potentially Displacing Individual Passengers Cancellation Penalties Often Difficult To Enforce Due To Competitive Environment -162- Problem Statement Given We Receive A Request For A Group Request Of Size S For A Specific Origin/Destination And Date. What Is The Minimum Group Fare An Airline Should Charge Given That We May Potentially Displace S Individual Passengers ? "Answer: Total Expected Revenue Of Displaced Individual Pax Min. Group Fare = Size Of Group Request Example: $ 2,200 -------20 $ 110 Per Group Pax ,.,111011, 1 -163- Group Passenger Seat Inventory Control Two Solution Methodologies Case 1: Assume Group Is Indivisible. Find The Itinerary With The Smallest Displacement Cost Of Individual Passengers. Case 2: Relax Indivisibility Constraint. Find Optimal Split Over N Possible Alternatives For Each Group Request. -164- Group Passenger Seat Inventory Control Large Hub and Spoke Networks Operated by Today's Major Carriers Allow for Several Different Routings (with Similar Departure and Arrival Times) For Many Origin Destination Pairs. For Example, Delta Air Lines Service Between: New York (EWR/LGA/JFK) and Seattle (SEA) Dept 5:2OaE 7:05aJ 8:15aL 8:20aE 9:3OaL 9:50aE 11:0OaJ 11:29aL 11:55aE 3:29pL 3:29pL 4:15pE 5:l0pE 5:2OpJ 6:45pL 6:5OpE Arr 11:35a 12 :3 0p 1:4 5 p 1:45p 2:45p 2 :4 5 p 8 :2 3 p 5:10p 5:10p 8:25p 10:40p 8:25p 12:25a 10: 4 0 p 1:33a 1:33a Flts 377/ 835 1429 / 1655 467/ 233 281 / 233 937 / 623 583 / 623 1601 / 301 983/ 833 887/ 833 1187 / 367 1187 1038 / 367 237/ 300 1425 / 1187 729/ 625 1421 / 625 Stps/Via ATL SLC DFW DFW CVG CVG MCO DFW DFW CVG 2 CVG LAX SLC ATL SLC , 14111 mi,, -165- Group Passenger Seat Inventory Control Numerical Example: Setup Dept Date : 12 JUL 91 Group Size: 15 Published Fares for Possible Outbound Itineraries EWR/SEA on 12 JUL 91 DL DL DL DL 583 EWR 623 99 EWR 197 950A 108P 340P 652P CVG SEA ATL SEA 1142A 245P 640P 910P 72S 72S 757 757 DL DL 887 EWR 833 1155A 312P DFW SEA 226P 510P 72S 72S DL DL 281 EWR 233 820A 1152A DFW SEA 1055A 145P 72S 72S Y $642.00 O/W B $425.00 O/W M $325.50 O/W Q $277.00 O/W -166- Group Passenger Seat Inventory Control Numerical Example: Results Itinerary #1 Seat Flt 583 Fit 623. Itin #1 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 24.40 24.24 97.20 121.60 93.80 93.10 118.04 2.59 20.83 19.53 19.41 92.99 19.25 17.72 81.42 78.50 117.31 114.58 109.71 103.29 102.92 100.67 96.22 6.21 77.91 94.12 14299 77.55" 14.92 74.68 74.19 73.15 6781 92.54 89.60 24.21 12058 12.42 11.79 11.51 10.23 Legi1 88.88 83.76' 83.51 66.82 65.29 Leg 2 86.77 85.57 79.60 78.33 75.52 Total Min. Group Fare Calculation => 1426.75 15 = $95.12.. -167- Group Passenger Seat Inventory Control Group Booking Model Output Rank Itin Out 2 3 4 1) 3) 2) Min. Group Displacement Cost Fare Estimate Per Passenger Request for Outbound 15 Pax Total Leg 2 Leg 1 $87. 80:. .8 87.80 7 .. ......... 4) $95.12 95.12 78.64 16.48 $137.93 137.93 137.93 0.00 $236.51 236.51 157.13 79.38 Lowest Published Fare for EWR/SEA on 12JUL91 $277.00 Minimum Group Fare Negotiation Does The Rest ! $87.80 -168- Group Passenger Seat Inventory Control What Is The Optimal Reduced Fare? For The Carrier: $277.00 For The Group: $87.80 Competitive Advantage Carrier Implementing Displacement Cost Strategy Has $277.00 - $87.80 = $189.20 Of "Competitive Leverage" Ilk -169- Group Passenger Seat Inventory Control Case 2: Relaxation Of Indivisible Groun Constraint Group Seat Itin 1 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Itin 2 Itin 3 Itin 4 259.11 255.68 247.17 246.32 245.43 244.42 234.76 232.73 230.30 228.77 226.53 226.22 224.97 224.46 220.83 139.99 139.97 139.95 139.92 139.85 139.73 139.52 139.16 138.55 138.44 137.59 136.08 135.31 133.84 131.09 104.97 102.82 99.52 96.09 92.52 88.85 86.32 85.08 84.85 84.46 82.47 81.23 77.34 77.05 73.42 Model Output Optimal Split Over All Itineraries Itinerary 1 Itinerary 4 5 Pax 10 Pax Minimum Group Fare (Divided) Minimum Group Fare (Undivided) $81.79 S87.80 -170- Group Passenger Seat Inventory Control Question: Why Are Groups Different From Traditional Bulk Pricing ? Answer: In Bulk Pricing, Marginal Cost Of Each Additional Item Is Non-Increasing. Example: 6 Bagels at $ 0.40 I item 24 Bagels at $ 0.30 I item But: 6 group pax at $ 175.00 I pax 24 group pax at $ 189.00 I pax Each Additional Passenger We Displace Has A Higher Expected Marginal Revenue Than The Previous One. The Larger The Group, The Higher The Average Fare -171- Group Passenger Seat Inventory Control User Optimal Strategy - Be Flexible In Times / Dates - Be Willing To Split Up - Book Only As Many Seats As You Need Carrier Optimal Strategy - Find "Minimum Displacement" Seats For Each Requested Itinerary - Try To "Split" Groups When Possible - Book Only "Genuine" Seats -172- Conclusions - Minimum Group Fare Based On Displacement Of Individual Passengers - No Distribution Assumptions Necessary For Group Passenger Demand - Given N Outbound Itineraries And R Return Itineraries, We Can Find The Best Of N * R Possible Combinations - Optimal Mix Of Divisible Group is No More "Difficult". All Necessary Information Exists ! - Better Utilization Of Excess Capacity Means Greater Revenue Potential For Airlines Ww''I ,,111 01, -173FLIGHT TRANSPORTATION schid)1mag LABORATORY, tt$ Wt 1@- Presentation at the MIT/FTL -Industry Cooperative Research Program Review May 23 /24, 1991 Professor Robert W. Simpson MIT -174FLIGHT TRANSPORTATION LABORATORY, MIT -- N Problem Statement GIVEN: 1. A fixed schedule of flights F for one type of aircraft - a flight is one or more flight legs - arrival / departure times are fixed - schedule is cyclic over a day or week, C - schedule remains in effect over planning horizon,H 2. A set of crew bases B where a number of crews NB are domiciled to fly this type of aircraft FIND: the cheapest set of work schedules, or "bidlines" b for these crews during H which does not violate work rules imposed by regulations or airline/union agreements; - a crew trip t consists of a series of flights to be flown starting from base and returning within one or more days - a work schedule b is a set of trips away from base on various days of the planning horizon, H -175FLIGHT TRANSPORTATION LABORATORY, MIT Typical Crew Work Rules - 1 1. Regulatory Rules (imposed by civil aviation authorities for safety) - Maximum Daily Flight Hours - Maximum Weekly (or 7 day period) Flight Hours - Maximum Monthly Flight Hours - Maximum Duty Hours (duty time is time without rest) - Minimum Off-Duty Interval Note- Crew trips and bidlines which conform to these rules will be called legal or feasible. These rules limit crew utilization to be substantially less than that expected by airlines from their aircraft, and mean that crews and aircraft cannot remain together during trips away from base. It is desirable to estimate the minimum number of crews required to cover one cycle of a given schedule as it would give a lower bound on the number of crew trips which must be generated. It is easy to compute the maximum number of airborne crews, but due to these constraints it is less than the minimum required crews. Due to the aircraft flying perhaps 18 hours per day, and a daily duty limit of 12 or 14 hours, some crews must start their duty in the middle of the day to cover late night flights. Due to the minimum off-duty interval of 8-10 hours, crews on late night flights cannot start flying on the earliest flights the next morning 1-- -176FLIGHT TRANSPORTATION LABORATORY, Or Basis for Crew Costs There are two kinds of crews: cockpit and cabin. The cockpit crew flies together for one month, paired differently each month. Each aircraft requires a fixed crew. The cabin crew complement has a minimum, but higher loads causes more members on certain legs. Changing reservation information can change work schedules dynamically. There are three components which determine the monthly pay of crew members at a US Airline: 1. Monthly Base Pay - independent of hours flown - depends on grade and longevity 2. Hourly Flight Pay - $ per flight hour - depends on aircraft type 3. Trip Credit Pay - $ per trip away from base - depends on details of trip itinerary - may be zero 4. Overnight Costs - costs of meals,food, and transport to overnight crew away from base MIT ". IAll ii, -177FLIGHT TRANSPORTATION LABORATORY, MIT 'N Typical Crew Work Rules - 2 2. Airline/Union Trip Agreements - Daily Flight Guarantee (eg. min. hours if called to duty) - Flight/Duty Ratio Guarantee (eg. flight/duty time > 0.5) - Flight/ Trip Ratio Guarantee (eg. flight/trip time > 0.25) - Maximum No. of Daily Landings - Deadhead Time is Flight Time Note- These rules may cause a "penalty" to the airline in the form of extra pay and hourly credit to be assigned to a particular crew trip if it violates them. The total flight hours paid in a crew schedule may exceed the number of hours flown in the aircraft schedule. Deadheading is flying the crew as passengers to/from base to other stations where their flying begins or ends. 1% -178FLIGHT TRANSPORTATION LABORATORY, Typical Crew Work Rules - 3 3. Airline/Union Bidline Agreements - Max. Monthly Flight Hours - Min. Monthly Flight Hours - Min. Days Off per month - Min. Weekends Off per Month - Max. Duty Hours per Week - Min. Off Duty Time at Base - Max. Percentage for Reserve Crew Bidlines These rules affect the monthly pattern of work for crews but generally do not cause extra costs. Whereas an aircraft may fly 300. hours per month, crews are limited to less than 100, so there are 3-5 times as many crews as aircraft. Note- Due to schedule deviations caused by weather, crew sickness, or aircraft equipment failures, reserve crews are given bidlines which mainly consist of periods when they are "On-Call" and must be able to report for duty within 1 or 2 hours. There may be a few flights actually scheduled into a reserve bidline, caused perhaps by holidays or schedule changes. MIT -179FLIGHT TRANSPORTATION LABORATORY, MIT The Current Airline Crew Scheduling Process Stage 1. - Generation of Feasible Crew Trips from Bases Stage 2. - Selection of "Optimal"Trips from Bases Stage 3. - Construction of Crew Bidlines for Bases Stage 4. - Construction of Reserve Crew Bidlines for Bases Stage 5. - Execution of Crew Bidding Process Note: 1. It is a sequential, heuristic Process and is not optimal, even if some of the stages are done optimally. 2. There should be some feedback of crew scheduling problems into the aircraft scheduling and airline market service planning. At present, this feedback does not exist since crew scheduling is done by airline flight operations personnel late in the airline schedule planning process. There is a need for some early assessment of crew scheduling problems in airline schedule development. 3. The availability and continual use of reserve crews affects the desirability of detailled optimal planning of fixed monthly bidlines. 4. A related process is crew re-scheduling by flight operations personnel when deviations from schedule plan are occuring. There is a need for good methods of solving real time, operational crew scheduling problems to minimize additional costs from disruptions. MEMEMOV -180FLIGHT TRANSPORTATION LABORATORY, MIT WA. Stage 1 - Generate Feasible Crew Trips STEP 1 - Establish "Flights", F For various reasons, it may be desirable to have an unbroken sequence of flights or flight legs; ie., there may be some arbitrary specification of where crew connections can be made. Even though their flight number may change for marketing reasons, here these sequences will be called flights, f, belonging to a set F. Every crew trip t will now consist of a sequence of these flights. STEP 2 - Generate feasible (or legal) Crew "Trips", T Since there are a number of necessary and desirable attributes for a crew trip, it is necessary to generate each trip individually. It is not possible to create a crew "circulation flow". The number of feasible crew trips may be of the order of a million for a typical US domestic fleet of 100 aircraft, and in the next step,( the selection of the best trips to "cover" the schedule), the solution may only involve a set of trips of the order of twice the number of aircraft in the fleet, ie., we are looking for the best 200 crew trips. Furthermore, there may not be much difference between the top 1000 solutions. It is desirable to find some ''efficient'' way to generate only the "best trips" as top candidates for a "cover" or solution. Thus, it is vital to find some new way to generate trips which: 1) 2) 3) 4) have zero trip penalty costs and good crew utilization start from a given crew base after a specified start time involve a specified flight or combination of flights overnight at a specified location WWI Ib 4111111111 Ili -181FLIGHT TRANSPORTATION LABORATORY, MIT Stage 2 - Select a Set of Crew Trips for each Base 1. All flights in the aircraft schedule must be covered by the selected set of crew trips. 2. Since each trip starts at a given crew base, the flying assigned to that crew base by selecting a set of trips must be proportional to the number of crews domiciled at that base. 3. The Selection Problem takes two mathematical forms: a) The Set Covering Version; Minimize [C. T] given constraints ~E.T21 H.T > B where E is a zero-one matrix where columns j correspond to possible trips, and have a one in rows if the trip uses the flight corresponding to that row where H is a matrix of flight hours per trip, and B is the total number of flight hours desired to be assigned to a crew base corresponding to that row where T is a zero-one row variable to select trip j such as to minimize costs Since the constraints allow the row sum to be greater than unity, deadheading is allowed and the costs include all components. b) The Set Partitioning Version; In this form, no deadheading is allowed and the constraints are equalities. The costs may be reduced to only the penalty costs associated with the guarantees. -182FLIGHT TRANSPORTATION LABORATORY, MIT Example of Trip Selection Trips, t 10 1 1 0 1 1 0 0 0 CIIlll C c21 Flights, fi fi f2 f3 f4 ........ = Cost =C2 +C3+ c5+ c6 010 =1 . 0 1 1 0 1 0 0 0 0 01 1 0 1 =1 =1 1 01 1 10 0 1 0 f5 1 f6 1 0 f7 01 1 01 1 1 f8 0 1 0 a 6 8 5 Bases, B 7 4 b 6 16 C 11 0 0 0 . . =1 =1 =1 1 1 0 718 ......... ........ 6 ........... .......... =1 =1 > 10 < 15 <5 Cheapest solution to this Set Partitioning Problem is the set of trips (2,3,5,6) With a large number of rows and columns, this problem is very difficult to solve exactly. With a few hundred rows and columns, there are a number of interesting ways to get solutions. If the lowest cost columns can be produced easily, and the lowest cost column which provides needed cover could be generated, good solutions may be found quite quickly. The trip characteristics which are desirable depend on the bidline constraints and the number of crew available. It might seem important to generate trips which do as much flying as legally possible in a duty period, but this would just mean more days off per month for each crew. It is always important to avoid incurring penalties from the guarantees. " WWIIHN111111910N IN, -183FLIGHT TRANSPORTATION LABORATORY, MIT The Crew Tree - A New Method for Constructing Crew Trips It is possible to create methods which generate any crew trip from a given base and evaluate it for feasibility and cost. Such methods may be controllable by the analyst in creating new trips with particular characteristics which can be added to the cover matrix as desired to obtain better solutions. An efficient method of finding "best" crew trips from a base is to create a labelling method which constructs a "crew tree" on the Schedule Map for the aircraft. This tree is rooted in the departures from that base, and finds the best crew routing for any flight in the schedule if it were to be flown by a crew from that base. The definition of "best" can be varied but maximizing the flight time achieved is a good basis. The tree stops whenever the daily limits of flight and duty time are reached, so that it describes the "scope" of feasible crew routings from that base in one duty period. The labels indicate the routing used to reach any flight and the starting departure from base. Whenever a crew routing returns to its base on some arrival flight a "best" crew trip has been found. The crew can go off-duty at that time. The analyst knows that for that pair of departure-arrival flights at this crew base a crew trip has been found which maximizes the amount of flying achieved. It is possible to extend the tree construction to find the second best and third best trips at the same time. Crew Trees can be constructed for each base. Best trips can be extracted and the next tree constructed to generate more trips. It seems possible to generate a Crew circulation for one base at a time if needed. -184FLIGHT TRANSPORTATION LABORATORY, MIT Handling Overnight Trips For discussion purposes, assume that there is a daily cycle in the aircraft schedule. Since there is usually a small number of crew bases and the aircraft schedule requires flights into. secondary cities later in the evening with an early departure the next morning, there are identifiable "overnight" visits for aircraft and crew. Since these overnight visits cause out of pocket cost, they require special handling. There may be more than one crew overnighting at certain cities. The crew tree will show the "best" way to route a crew from any base into the overnight arrival flights (if it is possible). Since there will be crew duties starting the next day at these bases, A crew tree is constructed from these overnight bases showing the best way to route crews back to base. By examination, it is easy to find the best two day trip for overnight crews. The search can be extended to three day trips if it is allowable or desirable. The selection of low cost, efficient overnight trips can be made first. Once they are fixed, then all other trips must start and end at their crew bases within one day. The departure and arrival flights used for overnighting are then removed from the Schedule Map before constructing the one day trips. The crew tree method is a new way to generate candidate trips for the second step of selection using some search methods of solving the set covering or partitioning problem. It is designed to only put forward the best candidates and keep the selection matrix very small. The process is not optimal, and it is intended that the analyst should be able to participate interactively in these searches, and return to this stage after the monthly bidlines have been initially constructed. 11 hil, , , ., , , 11,11dibl, WillivilliNIiiiijiillwk U1 114 INA11, -185FLIGHT TRANSPORTATION LABORATORY, MIT Stage 2 - Constructing the Bidlines Given the trips that are to be used for constructing the bidlines, there are a variety of techniques to generate potential bidlines which obey all or most of the bidline rules. These rules may be soft in the sense that crew schedulers may know where and how often they can be bent. There are research problems in beginning and ending the monthly bidlines, or"transitioning" between months. For domestic schedules, there are also problems arising from weekend deviations in the daily schedule. These problems may be handled by Reserve Crews, but if good methods of constructing bidlines can be automated, there are likely to be efficiencies in the number of reserve bidlines used (and therefore the crews required to support the fleet). One method used to generate bidlines is to create an efficient "pattern" of trips over 7 days, and to involve 7 crews in flying exactly the same bidline for the month. This reduces the size of the selection matrix. A much smaller matrix of trips versus patterns is used to select the "best" set of patterns to be used. The focus then changes to finding good candidate patterns for the bidlines. The patterns can be "mixed" to provide some variety in the crews' monthly work if desired. The solution of the bidline selection process once again requires a good heuristic search methods of quickly"solving" set covering and set partitioning problems. q%= -186FLIGHT TRANSPORTATION LABORATORY, MIT Conclusions 1. There are some new approaches to creating interactive methods to support the crew scheduler in finding good low cost schedules for the crews. 2. There is a need to create methods for re-scheduling crews when deviations occur in executing the schedule. This should affect the current use of reserve crews 3. There seems to be a need to create similar methods for the cabin crews which are responsive to their differences in scheduling rules. 4. There is a need to provide some "early warning" methods for market and aircraft schedulers to cause a feedback of expensive crew scheduling problems before the aircraft schedule is finalized. SCHEDULING SYSTEMS: COMPUTER AIDS FOR EXECUTION RESCHEDULING 1. 2. 3. 4. 5. 6. 7. Execution Rescheduling The Influence of Rapid Advances in Computer Technology The "Airline Scheduling Workstation" (ASW) A 2-Stage Development Approach for An ASW STAGE 1: A Manual, Interactive Graphics Scheduling System STAGE 2: Automated Decision Support Conclusions Dennis F.X. Mathaisel Flight Transportation Laboratory Massachusetts Institute of Technology Cambridge, Massachusetts 02139 USA (617) 253-1761 -188- 1.EXECUTION RESCHEDULING GOAL: - execute the operational schedule at least extra "cost" due to schedule aberrations INPUTS: - Operational schedule Operational deviations - Weather, breakdowns - Late arrivals - Expected traffic loads and revenues Short term operating costs OUTPUTS: - Modified execution schedule - Cancellations - Delays -189- II.The Influence of Rapid Advances in Computer Technology - "Techniques in search of an Application" - use mainframe: large, fast supercomputer - construct fixed code for technique - user submits data, receives solution - user reviews solution to comprehend it - causality: user cannot ask for explanation of solution - user may interface with OR analyst New Approach - "Customize techniques to the Application" smaller, interactive graphic workstations on - create various fast heuristics to solve - create links to solve large scale problems on - user ismaster, computer isservant, direct - processes are custom built to meet - systems to match existing procedures and common network subproblems mainframe interface application needs organization -190- Ill. THE AIRLINE SCHEDULING WORKSTATION (ASW) A COMPUTER TOOL FOR AIRLINE SCHEDULERS BASED ON THREE NEW TECHNOLOGIES: 1. Table top Engineering Workstations with a speed of 1-4 mips and disk storage of 100 -1000 MB working together on a local area network, interfaced with existing airline mainframe systems. 2. Large (19 inch), high-quality color displays with interactive,instantaneous, manipulation of schedule graphics information using a "mouse". 3. Object-oriented programming to provide modular code, easily extendable to handle time-varying scheduling constraints, policies, etc., and to reduce programming support. We shall call this tool the ASW (Airline Scheduling Workstation) -191- IV.DEVELOPMENT APPROACH FOR AN ASW GENERAL DEVELOPMENT STRATEGIES - Involve schedulers at all development stages-- (there will be cultural and organizational shock) - provide familiar systems and reports first to ensure that the new system will not preclude doing certain schedule subprocesses by old methods. - Expect changes in organization. and procedures as workstation capabilities are perceived. - Establish a local area network of workstations in scheduling area, capable of interfacing with the airline's existing mainframe system. (e.g., 3 workstations at $15,000 each). (Establish a "Schedule Generation" workroom). - Develop modern, transportable, modular, object-oriented software, for automation of sub-processes in scheduling - easily extendable - easily supported - C, PASCAL, LISP language - good data structures - A Two-Stag e development process - TAGE 1: introduction of manual, interactive graphics scheduling system introduction of automated decision support - STAGE 2: -5- -192- V. STAGE 1 - A MANUAL, INTERACTIVE GRAPHICS SCHEDULING SYSTEM A) Provide computer graphic displays of schedule information - instantaneously modifiable by mouse - global data base modification - selectable screen data -- by fleet, station - save alternate solutions - audit trail - memo pad for scheduler - keyed to input data, and assumptions used - automated search routines, etc. to minimize keyboard and mouse work B) Provide instantaneous error flagging (even if error occurs off-screen) - e.g., insufficient gates, flow imbalance, double crew layover, violation of turnaround or transit times, insufficient aircraft. C) Integrate crew, gate, maintenance schedule with aircraft schedule D) Provide familiar printed reports and graphics for distribution around airline E) Provide interface to mainframe data system to maintain current scheduling processes. ___I OTHERS .. -....... REPORTS EDIT FILES I INDOW SEARC INTERMEDIATE FUTURE WEEKDAY ] QUIT ALGORITHM ___I 4W OTHERS REPORTS EDIT SEARCH FILE .- ~... WINDOW R.OT A1T1ON_ .CHA RIT SCHEDULELISTING STATION CHART STATION PROFILE SOB PK0F4IE$ I EKDAY QUIT ALGORITHM 00 OF : M2 ne OF LAX NR is SEA: :2st LAX fe 676 B( PSI NCO : 221 ts ORD PBI CO - MIA: 26s Al FLL to; 96349/84 2820 SFO1 a : ATL 146 LGs L AX BOS s: ;. a ;1 494 ATL |PI | 8820 eoeo t39 ATL NCO: 12 t7 : ATL a s X 8998 7 LG A: ufLouf~p * FLL 6s t26 ENR 364 FL U434 | 2" SDL miamagp 44s |ATL ;ese9 CO FLL it ON001= NCO: too DFN 11 a 1 I " JFK LGA 1912 Los a FLL L10 0 ATL | a2455 a 63 BOSa 'us 8~56 egas 1180 * e63497e4 MCO: 22 7 S. 66 ATL 3s4: G~54 04,54 LIO Lie OF i 208 010301100 1s , TPA 221 S6 #: 0091 14 13 SF0 2 BOS 108 I 2 1 to e9 ENR so ATL us 006 LIO es I SEA 2010 LID er es LAX :261 ATL es An. s' DFH: 002 L10 2 0 4 e0 0 5. Monday deltanew ROTATION CHART GMT~e istF[ 3ss Al MIA ess DFN 14s F FLL 8B4 16 ATL :us : M( 9829 8898 es4 105 n2 JFK: 66 ATL El FLL 2" 016 NCO t2: ATL fss TPA EMR 0850 zu FL ist 9' 2 105 mi I EDIT ALGORITHM ADD COPY delta ni ROTATION CHART DFN: se 002 LI za z s :LAX OFN SEA ii ENR as TPA DI SF073s3 MO VE 1 MOVE ALL so SF0 O A"l 13431 SEA: SWAP LAX Ilre UNDEi.ETE ATI : :2s A 3 MCO 2f 006 MIA: L10 I - -- 14Z6 :ATL : 6s: : n FLL 013 LI9 iFLL ENR 364 2125 I DFN FLL MCO: tse ~yu~uysi tstF1 SF | das Al MIA 44 : 49 is MCO b U51 !q4 e44p :PO JFK: ATL tz s AlTi L FLL : MCO * ui~u~'~p TbY 2 BL se 17 | JKi 22 1in ATL U~34 I~bI34 S * to ATL 134 |PBI LGA is :12 1q12 ATL s3, FLL 2 0 BOS: LYOb :66 8p88 LGA: 011~in 434 * LAX BOS ~2 10 LGAs - S. 88: ' S 2 LID gl ss 'Its JI SFO: Al is U~34 NCO: FLL Is : BOS ORD ATL 3s4: Io' 007 :sis B( PBI BI : ATI : 221 its 22 6 005 LID 015~O LIO 010 15 77-7754 1-04 012 LID92 14 13 12 MODIFY LAX 003 ggg 008 0 11 10 09 INTERCHANGE AiTL se* Monday DELETEALL 0 04 2 ~8188~ DELETE 21 143 e ts si : , : ATL| II UF is M( ATL E) FLL: J. 011 feiL 761 MCO 123 ATL i ss EHR TPA I 08 'ot ' Fl ENR Ii SLI I MIA NCe6 6139 9214 CC A 394 of6s34l 45,14 ''no 9549 225 : LAX Ld 9p U . 66 '969 9~4 6899 III SI P ATL 4 LGA SFO -FLL S 36 S I ATL IIII SI ISII II I I I uso FLL MCO II I S FLL 0 01 0 07 ul 767 se DFN: 1 DFN 002 03 04 05 s1: AiL Is 02 07 10 09 L0 ENR gg 13 T: iPA : : 01 I 1' 1Op56 AlL261NCO 0900 : a2aMi tiit 10B tt. s BOS: _ _ _ _ a a a i-u El 9 SJU :3 ATIL 7s 96 Al ORD 1 : BDA: ATL is a a3t apa LOSAX ( 4 aT6 arw Al %6 a a a ATL___:___:______MA_____ pau'.4 tsi:MCO :112 AlL F :i: LAX Al t46: AlL:is% LAX: FLL 77 DI 1 __________4_ sas DF 16f1713 ATL: a2 a ao LGA: 1 51024a SJU 1155 _ _ Ait S F O m s 008 : so ol _ _ _ NCO 205 01 063fo4 5 :NCO: a a 072gsamluaa :ss 3s ATL 1HIA: L ATL 11 _ _ ORD13 s :u 1 1 _ PBI u6 I LAX 1 4 1122s BOS tit s 2$ AIL NCiss 14s 1E23 PBI ' ATLisv 22 21 DFN as :AX a 20 1'~ so : a 9 1o8 as' ie I SEA:01 132 1170 17 16 !~ SF0 2oIo 007 I5 14 2 005 Oil___00__4______ FLL 012 118 0s 11LAX JFK LGA a6, is: DFN MCO ~: : ts aL 1 a a aBD 015go JFK: n21 a aa asMOu a Fl 151 :AlL its a PI a13a2d~~ AT I I :NCO ag 1a JFK 144 Al: 7s MIA s Al :443 a FLI :176 L1 110 a :ELLamCO 01L 04 364 ELI ENR 12s is01 0 12 z17 SEA 16t: ATL 010 11 ENR LI5 004 ALGORITHM QUIT WEEKDAY Monday 08 :LAX :LAX:1-43 n 003__ 0Q I :s2: WINDOW FILES SEARCH delta-new ROTATION CHART GMT EDIT REPORTS OTHERS isi 4 Al: 1.2 A si so aAlL: :1221! 52 IBDA: ;IIl Al FLL NCO 62' :I0Ss p :14 76t :NCO 123' ATL ass TPA a 0400 a1 2a a Ep0 R *b FLL 69 ENR:829: i 0 : ELI ENR 2 16 ~ A i t 05 FL :Al saJF OTHERS REPOR" delt a n ROTATION CHART so L02 011 : DFH: :LAX Monday Monday DELETE DELETEALL INTERCHANGE TPA asa ATl tie SEA: L04 2 1A 1 2 2 NCO aI g 576 : OS :ORD is AlL: I 23 2,127 ATL: 15o a FH I I 91 tit: NCO tit ATL ll :s 22 M:IA: F ~ ATLU6PI NCO: L 2C2 0080 iLs A LAX ri 96! B:OA: 61 Al 66 :S a : a : :i5 :MIA: 97 ATll : LAX :Lu Al isi:NCO :tz AlL : %6 as aF LAX: Lm NOS IGA:n AL:34hl: ae TLt6TA :FLI 0a :134 . a a 4 : -MIA m21f aTANa 085 aw Al UL is% Al: F1 isi AT:t2NCO 14 : In .FLuM4 ____87______6s_:__ a B t 46: a, a ORD 63P ; AT 739: P11 ins :Al ?4 JFK 144: aLLlaC a 78: l aT C2AL2 a a a ape a a F 7t 110 :14: ATL Al :32 C0 011 00100 015 1 :10: Use SJU : 66s :494ATL PBI LG6 2 : EE N: 22 7 10 I3 010 ' FLI 2 AL 394 :Lf t~ SJU :33: OF tit L05 106 24 o DF LAX :s a a I LLUb ZOM:PB1 st NCO 2st ATL 119 is :LAX :t~s LINDE LETE EHR 107 1 DFN: SF MOVE ALL MODIFY SWAP SEA 70 ALGORITHM QUIT WEEKDAY WINDOW FILES RCH ADD COPY 07 :z : LAX: DFN L03 000' ENR : Ail 03 , I1 1 EDIT I'~ a7 1~a REPORTS OTHERS oo -7 0 1 so DF: sos DFH AL 07 50 08 Monday 0 :9 10 L0 03 ENR 0 L t SEA: ORD MIA A3s4 ATL NCO2si L0 722 10a FLL 2 LAX OS ses n s : ~a : ATL :194 2756 0eyUWv4 pyIu JFK :JFK: 9 Al ORD 7: B( ?4 ATL os ATL : s l LAX ul 134' :FLL :5t : FLL :364 ATL too MCO :44s sas s L( j : Ft i PBI 14 A:TL t@2 176 : I ATL :u1s : NCO so ATL ATL L s4 ENR : S a a:E obt ENR FLL BDA: es aO85a11~l ~u ATL : ENR :25' .".5'1~qu~'q I o :JF FLL :87 Us : U0Ssa 62' : a a MCO SFO 4: :gi 01 2 a'2 : :1 ~ Al -JFK a ATL fss TPA 2- t46 ATL: : 144 :ATL iss: DFN ts F49LL IC a1geg FLL n;i:F F MIA 2vsoosi CO 12W B if 3 1pe EDL 761: MIA 6 a All TLs:MCO :112 ATL LGA e32 e5 BDA: 61 '2 DFN Lf s, ATL :2' LAX 126 : ENR 13o: BOS: is FLL 76 05 SJU 66 4s4 ATL PBI 2355 LGA: 0116"U LGAs SFO: L ATL itu sm1 2zgzmmm 90 NCO: FLL ot ' 7 BOS DF I aa 2 00006 71 SJU :03: ATL: is 1 5a.24| 2 - e s NCOtit ATL BOS ut P:I :67s ATL :ts: PII: 05G DFN to: LAX: 2 , ATL: ts 1320: ,134287 II43I ATL ist MCO st NCO ATlliss so SFO so DF : LAX is :DFN: 24 23 22 21 20 19 1 17 1 15 TWR iPA as s@ ;ENR: SEA n11.1 14 13 12 11 :LAX :LAX ALGORITHM QUIT WEEKDAY WINDOW FILES SEARCH delta-new ROTATION CHART G-----a T EDIT i- ALGORITHM QUIT WEEKDAY WINDOW FILES SEARCH EDIT REPORTS OTHERS M -ROTATION CHART 7 I delta-new I 15* g 2 I I DFN:M 9' 002 g og DF L1 : A& SF SEA: - 92 : LAX SEA: :LAX : nt : A LAX 003 FOTPA :i EN R 04 EHR k6 : =I.1 2T L u0 2 : : : SF0: tit AT s LAX OF ATL: V D NCO 1 tit ATl f. : OF AI s OS: a 8 26' :ATL :94 BOA: 61 LB MI AWs a s: A"T :4s4ATL PBI : : NCO ORD : t 34' ATL: BDA:62 86 so'Al: ICO FL NCO SF0 t 49 143 Is% UL ATL: 162 14 :Al : JFK 144 PBI ATL 7ss MIA ins 14 t 46: LAX FIL LL :FLLaMCO ATL:. ATL si ILI : Lis 6:ENR :Al ATL :s MIA 60 ATI tsi:NC0 :In ATLl %6 : 3 :SJU : :s3 :66: :364ATLNR44 t Os LAX FLL 0 1TDFL : :LGA; LAX 71 o ATI is :676 7 IA:W ATL 639F4 10104 2 01801081: 13 I s SJU OO~1gsi : 57 RD SFTL ALlseg BOS GA: ATL s@@s9I' ORD: NCO261 FLL 7L ANC tot lOS 11 08ggiggM~i I ATL is1? CO 002sMIA 116 P3A: L06 I DFN: I1524: I 715 t ay PB I u s iLAX~ tis A _ : -_ *TPA 1 0 ' on IOSs 6k D2s t21:FLL JFK: 01 61: COZ 01066 LI 2 ATL "Wsp ATL fs6TPA : 2 :: : : I : I :ENR: I 0 obt :194 FLL : NR es memas : u' : FIL E R:s2s' q esy 66 JF |ATL ,ggl I tua a OTHERS IRE POR m w T 7 n c es9 As DELETE DELETEALL delta_n ROTATION CHART I A 1 9 2 TPA NR: Monday 17 un L000108' SEA: 0S ZOOM ATL: 10 0010 007ggooio 10 79, ATL MIA: s MCO 261 06 SF : 008 100 ATL t6 BOS: ATL L1G300100 f13ooioo FLL ATL FLL aATL us | NCO s' ATL ATL |A1 | ENR FLL 6ms us 6 Oo10 ENR sat g80 as__l ~n JFK 14 PBI :ATL: 79s ?ib SEA ENR |s2' gg it ATL 2 A TL ts PlI ORD FLL tos BOS 14 S ses ATL BOS 2 LGA ts FLL tss NCO FLL n7 :12s 7s ENR W- LGA a BOS sk BDL |s JFK: 12a :NCO 128' ATL Cs6 2a" A' FLL g st 9 7: ATL: t@2 4 a iBDA6', s ENR 146 ATL 14 sis a LAX s 1re ist : t ATL ATL 9 :LAX tra ss is MIA s44 ts ATL : DFN s s SJU LAX LAX FLL :364 ATL aa I s' DFN: o tit ATL S5 2 t JFK s6 TPA : ~~ aCO esag0e 011 O16@gg : MIA: 6 r ATL t97MC? iua BDA: 63 spee1 L009001001 94 93 1 55 s2 SJU 4S4ATL PII :194 LGA LAX ;s its aCO: 02 01 s: ATL: DFN U LNDEiETE ORD: 24 23 LAX MCO its : MODIFY SWAP 26t NCO L00188 : s: ATL iss e LIG 003 22 21 MOVE15241 SF0 002go, 20 19 is DFN INTERCHANGE 0 ALGORITHM QUIT WEEKDAY WINDOW FILES RCH ADD COPY r1111w :ATL |76t gg FLL delta-new ROTATION CHART MCENR: 01T 002 s SF0: ap 0 43I SEA: t : it: ~ BOS 67s L10 NCO 2s1 MIA: s '0 L06 Ibq6 ATL 2st EHR26L ats NCO: SF0: 007~i LGAs 4s4 ATLPI B0S :s ATL:s:PI 2 ATL I:4 is sSf 2 12 :1 ATL ist:NCO 112 ATL LAX LAX ATI: st 146: 2 015 14 Lo16 JFKQ Li80 0 FLL a ) ATL s ATL 3 FLL 176 ATL its N FENR6 ENR 1:4 ATL LiO esATL: T ATLiss FLL' * sit 085 100154 ts2 -FI LGA 76 FLL EMR :2s NCO n FLL S I |NCO ATL 1 PBI ss DF FLL LOY 016 80180i MIA ti JFK :AL-s 14 ILA FLL :: tst BOS m __1 93 &pUa O'e : FL i BOS 24 ATL ATL MIA2 s' 6s ORD ATL 2, BDA: s1 2 : -4 IU ; 01* J SJU 66 ATL SEA: i :LAX, 705 0051601881 10 07O LAX , 1715U1 F'1: ATLf41 SJU s3 ATL: is tt NCO ATL 1;u u1Y : 05J1 v5 ,2 DFN tr DFN : :152 :1~ m s DFN: 1 1 s ATL: 14;55i~;p 31 128 PBI NCO ORD: mCo its LA VG HI 3 ~ 1-40 03iO 261 ATL iss' ' :LAX s LAX as: Monday 04~ a3 82 a 24- 23 22 21 20l 19 in : DFH: '5 i TPA a * 7 t a. EEDAALGORITHM WIND FILES SEARCH EDIT REPORTS OTHERS I* BOA :Fe:' E4 BUS 6k BOL 62 FLL s@ FFK: :463 ff1% 12t, FLL AATL 1A5s N j 2'EIT 2MM OTHERS ROTATION CHART GMT 05 06 07 08 s EDIT REPORTS SEARCH 10 11 12 14 * lie :66A 00 NCO: 24 DF~i 02 01 sin: l t~. tit : :IOS: s1 poo f 04 ATL so' :DFN: :FH 05 LAX: s2 SEA: 70A~ I ' aL1--s Sa a fi 26t ERRLX I : DF 1 AJL s6 E a LA ATL i:s P aT1 153 is: W J tr .a R 24 ATl I :105 Lot FLL II a 2945.Zpe 0 08 @pl@6j - : LOS9881883 009s Bew AL i:NCO|u2ATL U2 AT1 LAX 10 p a a a gs4 a,10 er vV LAX ; LAX JFk L1 : 02: FIL :364 AlL :443 :FLit.MCO :151: MIA sis : :134' FLt. 144: AL: 7ss: PB1 ATL: sj -iFL FIL :ATL: t#2: NCO NCO sAL is :A: :XATL:its ENR 126s F :s LS 6 a6 ATL 015 L~~ 1 LI Dmmn ' a 'i c IPA a a : : : : a a ' ENR: a sat 194 ER : FLL tes ml10 8t amm] FLL JFK: i2 F6 EN[R 54:s32' ggmaiamip' 10 :ATL rut C FLDLL 10 010 IGA: :76 l s' BOS 2 tsSF0 FLI1 ses is% ILA JFK 14 19D ellees 211: ATL 46 eYev 0100:1:0: 0 6a 03 171 ATL :2' 1 BOIA: ' NCOt AW DF a :s: SJU :lp ; esagr4 its 132 AT SJU LID L ALGORITHM 1 24:~j :12:15 IA: 3ATL 154 23 ts~vN ts:A :LAX: AL 10~ 22 LAX O56 005: MIAaa 1008 21 :s PIILAXa s676IOS 113y297 NCO 21 L10 20 19 DFN is 1 LjO4 2 17 Ao TAL 0:SEA: 003 16G .15 ,,SF0: LIO %;E is QUIT WEEKDAY Monday I3 TPA E:R 0, 143 2t MCO LIAD WINDOW delta-new 09 0 004 FILES MCO 123' 280 R"A 2F a as FLL ATL 2P 156 EDIT ADD REPOF OTHERS 07 01 0 ,o deltar Be_ _ : EN TPA R e eMODIFY * zoomET ZOOM24 261 MCO LIO tPE3 01861 ORD: is C0 MIA: ws 815,1 26t 8 0000 L00 LGs SF0: SJU . SJU 66 ~ ; uls NCO: LID : 1105 ATL :1s4 DFM * 52:~ ATL :7s :2' 715 ATL 16t 2 1t ORD I is IDA s7 MIA s r2 u SEA na :LAX: 01 : LAX 2 : ENR i ATL NCO tit 16-1713 9V ATL its PBI 96 ATL BOS 1 FLL ti: s ATL pe13 2 55p LAX %6 6 6 | 6: 1 6,up LAX 6 102~gg JFK FLL :FLL :364 ATL CO :44s LAX uW ATL is|:MCO :ii ATL ope65 15 o L0400 DFN el o11 I~e 6 ! BOS: 494ATL PBI tit BOS , ATL ATL: * SWAP 06 NCO its :"wrli ATL iss MOVEALL SEA 00 0 1 et MOvE1 002 010 At se' DFN so LAX ,ogFH:es TERCHNGE SFO 787 1-0 Monday DELETE DELETEALL 12 to_11 9_ ALGORITHM QUIT WEEKDAY WINDOW FILES C OPY O TH E RS__ _ _ _ _ ROTAT ION CHART - ARCH FLL is MIA ATL 799 ess JFK 144 PBI 4 145: DF1 f43 t12 SFO NCO FLL LGA 7s EHR t26l DFM tit BOS n7 FLL iss ATL: 14 2 LGA: isw ATL: 134' ses ATL 46 tes NCO 4202 FLL 6 TPA ENR 08650 sI ATL: NCO s76 ATL is ATL |4s sbt | 1 FLL :IDA: s* ENR 6es 62: FLL s' EMR s296 '1OS BL so ATL is6t JFK: 12i FLL NCO iws' ATL is6 20 266 42 EDIT REPORTS OTHRS QUND FILES SEARCH Monday delta-new ROTATION CHART GMT 5 6 7 0 :9 0 1 11 11 14 5 is 19 :DFN: 20 ATL iss 3 10 0050884 s BOS 06aai 15 LAX 10 W: BOS 82, SJU __ e05s3ge4 FL 0192 10188 I 14 103 ATL 364 Tulv MIA 1 ;H1995 - t6 ATL us tOsA LGA: u a TPA I g ps0 JFK 14 ae 1:, ATL: 1e2 NCO q91 SF0 :ey~5~ NCO so ATl ' s :BDA: : 62' :ENR: FLL n 1 is FLL iss 4: pa1 LGA :1s6 ENR m Ifr DF isi NCO iso 42m ae mnamnm. :BOSs6 BL aa 60 1I5 L3 ATL iss: PBI ia *2, - 4f0a%1 Is DFN 14: ELL 10 449 BOS ses ATL : 2 ATL 144 ENR ATL |s4 : sIm :sm :FLL 6es 1 Ua FLL ltI SF 1" 11AL4: FLL ist JK :FLLiMC0 , * L* 3 If BOS ATL 97 i I I 2 1 LAX 1:1 LAX ATL 2t -1 26t ORD ATL MIAA u: ATL 7s ATLI 10 o 81881 II EMR 14i ATL tis PBI passumiggi2 55 :07 SEA n0 ;LAX 1e1Pu24: 151713 ; : :JU ATL 6 10 n: t :MC ATL %6 :LAX s2 I is BDA I PBI As4 ATL pe 3 ss 007 LGA 0 715 NCO ii ATL : iii SJU 66 l CO DFN 524: : II 0 A :TL : MIA: 3s4 ATL MC026s1 o4 ATl s' :DFN: ut DFN :p L0 OR: is o2 03 1 - AIL T a NCO its : PBI L04010 "' 24 s: LAX : Arm NCO 23 12 LAX: LX i 22 1s 0~~(43;1327 SEA: 21 1 ' 4150 so gg SFO * * * to * .734 L10 t7 :: TPA is sI ENR SFOC THM FLL :v se :JFK: S-Uu aaIs.g2 ENR 15lig 32' :ATL: :t NCO AT 123' 1 umsamiINppNa umu p ;gga IIbIa FLL in2 2 2" 9 8 07 GMTeG Monday delta-new ROTATION CHART 's TPA I ER: s: tie 0m L04 73' as ATL MIA: NCOs 09 2 %6 LAX 6681 L* 10 4s ATL PBI LGA :SFO: : BOS : £s BOS :134 ATL Al tsiNCO:tizATL 05Y0 ,q/W MIA 65 :OS : 74 ATL FLL S0 s ATL2 iss FI OLL t7s ATL s JFK 144 Tau :15 s TPL :2: 2 ATL: t@2 4 PBI LGA FLL: 7s isi EMR 12 NCO 5u NCO s IU3.5 ATL 0 FLL iss ' MC:01 I 1 t3 IGJ L L0 BOS 22 M. 2N.1 I I L1 ses ATL: :46: 0 O12p @8 : FL n SF . MIA , LGA: _. I I ATL it ORD I I 134' :1ti iL0e SEA: 261 ENR is ATL: 173~~e JFK nt ATL 7 LAX 1 FLL :364 ATL :441 78: LAX 0 1 1 is BDA: :LAX, Its PBI : ATL : SJU : 2 ATL s 4 ATLm :LAX: fi 5 a s DFM, DFN i7t NCO tit ATL tut SJU ss as NCO: DFH s ATL 19 ATL: tv si; ORD NCO its : s: LAX ; :8 ATL iss LAX ziNCO:P1I 20t~ :DFN: I SFO SEA: s 5 14 1a 2 le 82 0 23 24 22 29 2, 9 0 t ENR 1 s'L :t4 " FLL I ZI ENR | s? Iu SATLF e BDA6F OS6 BDL FLL # n1" 6s s ENR ;AT JFK: 69 11199' 1t, FLL 2V C' t Ts76A MO ATL C3~ 91 I I R ATI N C delta-nI ROTATION CHART GMT 05 06 LIO 0s 10 iPA Monday DELETEALL 1,2 1 ENR: 700,plw 67 002 e7 09 :DFN: INTERCHANGE MOVEi MOVE ALL 21 20 19 18 17 261 ATL a DFN -- 00 0 008 2si 10 ' MIA: 3s ATL SJU A NCO: tie BOS: LOSS LGAs SFO: 4s4ATL PRI :14 s LAX us LGA: is% 134' L10 ATL 146: 01301100i LID i905 a :FLLfiMC0 :443 1 ATL 74 lOS FK is : 144 FLL it JFK ATL SF 1OS se 193ts 24ags0 LAX 364 FLL n7 : s LAX t: 010 s a DIN :ATL: 7ss a PI 14 49 is FLL LGA : ATL: te2 NCO SFO isi a26 2 919 FLL ass J711W aj~ MIA 2 1u 2 %s ATL is :MCO :i2 ATL : FLL 225 ORD epe:1e L11680101 ATL ist 26t ; t16 PBI ATL si MIA 6s " SEA mi 1 ATL: is ENR fnt ATL 131 BDA: ATL py 2WS :LAX 164u71 ATL :2. LI0D 10 008 36 pa :LAX s2 1 MCO ithATL it BOS SJU 66 : ui r tiszessu145 ZOOM ORD se DFN ATL U NCO its ss 1 5 MODIF Y zoo ATL :1415 SEA: L410 | DFN sa LAX s 04 03 02 01 24 23 22 o9' SFO ALGORITHM QUIT WEEKDAY WINDOW FILES RCH ADD COPY DELETE REPO R O TH ERS 225; 2 ts NCO ENR DFH 2 21r, aLL L76 NCO so ATL ATL its s 100 ATL 56 TP 0160886g L1O s444 :ENR: sh 0ga850ig BDA: RFKl :3a EaR a : FLL %is 62 :OSs k BDL 6s FLL EMR 82s' ATL :7t p120s 15-gO 121, MCO tra FLL ATL ss I ggp '" A r] 21P1" A CH pO delt: NC : e Y : OARD, s -10 1 s its PBI ATL Al ATL isA lCO ts : DFN :L AX z m v M q1 ROTAION HARTdelt-newMonday e E 2 as A4 es 23a 24 2 1 22 ATL 2S :SJU 6TL s: SFO SE 02 17 Is76 I QU WEEKD A WINDOW FILES SEARCH EDIT REPORTS OTHERS zetMCO 1 ADF s6 LAX s : LDA L s sAX AS 05 003 a 00100, L1 L1 at 00 1 L1 Iw 2a 5 a a4 a: em Iu MIA 11125 A. :G: : Al: 1~ OR 7@ ATL 8si 16:LK : : is LAI s4 :DN eATL:1: NC S FLL 44ATL Das IA 14 14''~ 12" a *6(ws A: aPB $ss pmma a a aI :' a53 a a a a 9 a a * 55a a 12 L1 aUl a0 avo a a0 13 tf1C i : aA 14 : Ta :443 364A 144@P "JF"!% iHCO sLX a i6 1. IU 1ipeT uaa 'FLL l isM Jk3FFLoL' :17s :ATL ts L a a a s: FLLes ELR s' t DFNt49 MIAR: ER iD MC3 ALsF !a" a a a t4 I 2af ' LAX iFL 1 ;Me; Ni " am a1 14 :3 anINC8: 364Al a a aA ATL : 39TLP4 SF026 LG A L1 ~ aU a 1 ATUU 731 L1C0 :A i ss :I E1R AT: TL FO 82 i FLL BDAUL D ~ ~ s : AGA : A s aL 2W'T7 w oo ON CL : 6 CO 2 : 2 is M L 26 FLLIn AL 5 i4F17~1F I 97001 06166" ~aC 1 2~. tsos 6 5 ORD: sEl4R: 1A s LAX ~ MIA 3s 814 ' _ ATL 6j-.3?rfml tis 25 2 19 SJU : _ 110 sl : FRIDAY FM 44 ATL PBI LGA LAX 0a aqe %6 ATL is a aa :MCO : 84 ATL :14 2eWEKL LAX as : ATL BOS 14 LAX a- a LAX :34 FLL tt : s ATL i tuATL ATL se ': i:s NIA: s7 si ATL ORD ATL :78 _ _ SEA ut DFN: s LAX :290 _1I8 n : ENR NCO 1AU Ali lit: BDA: 6S :LAX oz 1-10 SF0: 0 04 03 02 o SATURDAY SUNDAY s NCO FN :0s : ss BOS 6s7 66: : __ NCO: L~9100 ALGORITHM Monday THURSDAY OF as' TPA 3 21 to PBI NCO 261 20 3M i e 261 NCO 006010, 3 7igg 8 ATL iss s 043; 003 pjg 1 SJU SFO SEA: 1 17 a ATL 7 ogiga, 006 00100 QUIT MONDAY WEDNESDAY delta-new 02 1-05 WINDOW FILES TUESDAY ROTATION CHART 00 SEARCH EDIT REPORTS OTHERS I, a.. a- a- a~ I ATL: is% FK ts LGA: BOS sem: ATL 46: SF 21 n FLL 150 FLL ist 008001003 FIL :364 ATL id 04st Lo'00100,1 DFN a 1 ATLus a NCO ATL 244 :ATL: iss: PBI 14 FLL :12s ATL: @2: NCO 1m q.;s349 1 a SFO lse : 6Aa 87' FLL a LGA 7s ENR DF C its a FLL 5s TA 6'3 10 0 3. 14s 0FLLu(1C0 a1a a~~ IA M443 1 31ENR is 'ATL| ENR |14 aI FLL ENR s2s aes 112I""= 144 Z @" 2 BDL OS6 :FK: 6e |ATL 1% |761 NCO 121 12' 2 FLL ATL 6 2M delta-new ROTATION CHART 01 se : : ER: ENR ORD: 7s' ATI: 79' AL: ORD: :: 4 : 91 SJU :36: SJUJ :s; 116 PBI: :ATL :ATL %6 PBI: b S.. i~ _W__ 04 - a - 02 :7-SJ6Jis2A222 __s' 4ll 15-12EHkRD Eu MI,: R: Tu pl1: AGOIH UT WEEKA A WNDOW' S SAC EDT REPORT OmHERS I av Th glg detae Th soe : Al 79' ORD: NL s Raa D a : a s~ a aeekl : sJUi ATI :6: sAT J 11,1 ruT: B i Fr glo se ENR:ALSU6 Sa Su u ATL 9: Rs.:ENR: ER OD Su~3:N a Tu a aD a a SRD a T i PB a a a a 00 a S' ATL s ATL a He ATL : ~SJ so a IV if All OWs' . ATL:L :ATl:: :ATL ATL: IVI 51 N 02 a a :LA aB a ATL ss2~ SJU329 a Ai 96 :s6 CO:36: AL s NCO e se : IIJ to NCO ALliss'! NCOits F0: :so ALliss' SFo: ss S o As Ai TLissA MCOis6 D a0 ASF TL a SJU oti if ss~ DFN 109:1 :DFH DFN DF s2 :LAX: :.2 :LAX: LAX :LAX: :2 :,2: ,7 ThI Fr = SF0 se5 ATL: _ 1 0 m ts AL eDF ATL s 7t5CrL Sa As SE TupooI ATL iss NCOisi 0L DFN 32:, L LAX A a JE 4 13~i OTHERS - ELI 110 U fREPORTS ____ EDIT ~-... ___ SEARCH WINDOW Fit ROTATIO N-.CHA RT EEKDAY QUIT ALGORITHM 3CE UEL. EQ FROM 0_ DEE ARgy CLASS MEAL EE STATION CHA RT STATION PROF ILE L10 XG FLL JFK 1720 1948 8511 62 L10 XG BOA BOS 1650 1757 851110-999999 014-014-014-014-014-000-014 62 L10 X6 BoS BDL 1844 1925 851110-999999 014-014-014-014-014-000-014 63 L10 XG BOL BOS 0700 0744 851110-999999 021-021-021-021-021-000-021 63 L10 XG BOS ADA 0855 1155 851110-999999 007-007-007-007-007-000-007 65 L10 XG ATL MIA 1012 1150 851110-999999 008-008-008-008-008-000-008 66 L10 XG LAX ATL 0000 0534 851110-999999 009-009-009-009-009-000-009 66 L10 X6 ATL sJu 0704 1105 851110-999999 006-006-006-006-006-000-006 67 L10 X6 ATL FLL 1326 1505 851110-999999 023-023-023-023-023-000-023 70 L10 XG ATL ORD 1644 1735 851110-999999 006-006-006-006-006-000-006 74 L10 XG P1I ATL 1420 1600 851110-999999 012-012-012-012-012-000-012 74 L10 X6 ATL BOS 1705 1920 851110-999999 007-007-007-007-007-000-007 76 L10 XG FLL LGA 2025 2300 851110-999999 011-011-011-011-011-000-011 77 L10 XG LGA FLL 1955 2240 851110-999999 010-010-010-010-010-000-010 79 767 X6 ORD ATL 0815 1100 851110-999999 001-001-001-001-001-000-001 80 L10 XG MCO ATL 1105 1222 851110-999999 014-014-014-014-014-000-014 60 6UOSPRDFILC s 015-0 15-015-015-000-015 .' ~ -- WINDOW ROTATIONCHART F SEARCH - 05 881001 - ALGORITHM - CHART: SSTATION | |Saturday QUIT EEKDAY SCHEDULELISTING ---- TIME ZONE : ATL DATE EDIT REPORTS 4'OTHERS STATION PROFILE SUF PROFILES LOCAL 01 00 02 03 04 05 07 06 08 10 09 11 12 13 15 14 16 17 18 19 21 20 22 24 23 iL 3:9 94 NIa PI 176 112 1933 98 SF8 FLL iCU Ell IF 354 ilI 64 FLL IC LAI 115 117 17 Ili ili 449 P 799 Si8 Ell 194 MCI 195 SF6 97 Iit 31 07 08 09 10 11 12 13 14 15 16 17 32 El 116 SF8 96 SJi 146 177 LAI ORD 123 EC8 2619 Eli I + M4 IC 761 Lil 187 I 8 IC 1 155 06 111 NCO 1 L 83 sit 33 )5 16 LC + 44 4'1 4-4,4SF3 137 97 134 LAX 158 NCO 32 SJi 36as Ell 886 98 PI 116 It 96 TPA 156 ice 261 141 18 19 20 21 22 23 24 01 02 03 04 05 GMT l\) EDIT FILES EARCH ADD OTHERS WINDOW WEEKDAY QUIT ALGORITHM COPY ATL DATE : 881001 b ELETE - 05 TIME ZONE : s:iLETALL |Saturday| INTEPCHANdik MOV1E AU-L MODIFY LOCAL 01 SWAP 02 03 04 05 06 07 UNDE LETE 08 12 13 14 15 16 I 17 18 19 20 21 22 23 24 goo 1341 LAXI j354 I[i 64 FIL 57 1Ii 41, 14"4, 1 151 IeCI 32 Sig 1 4 4] 4, srI I MIb LAI Ell PuIT BP SF6 137 7 115 110 755 33 4 - 4,1 + -41 445 194 135 33 57 06 GMT 07 08 09 10 11 12 13 14 15 16 17 761 17 Ei[O go 156 116 6 L 155 123 Ied 26131 Eli IF IF cd 3 1 83 56 Sig 146 171 LAI Sib -U, 4-+ AW? 4, Sao9 05 16, Lgi tt 1 3 116 ieC EU SF3 -4, 44 I5 Ali 261 96 55 141 18 19 20 21 22 23 24 01 02 03 04 05 OTHERS DATE : 881001 EDIT SEARCH FILES WINDOW WEEKDAY QUIT ALGORITHM - 05 TIME ZONE : ATL REPORTS |Saturday | LOCAL 01 02 03 04 05 06 07 354 Ili J64 FLL + 08 09 10 94 FBI t76 112 73 FLL NEIII [1 + + SF3 137 115 117 7 449 12 13 GIGice 14 15 $:s IIe 83! 194 18 16 LGA 32 I 11 ECU ![1 ~j Il~~~~iijEI 57 ii*[ S0 SF8~~ SF3 97 Sig 13 83 I 5 19 20 IF 31 I 1 L I 21 56 SIN 16 146 177 F I 1 1 S9LI IIi %L11 B to 799 195 17 EI FBI Egg 16 134 LAI 15 I& 11C LAI I 11 22 23 123 ECU 261! l s Eli 98 335 336 FBI 116 oil 96 761 Lil 137 24 TPA 156 NE8 261 141 6i7 06 GMT 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 01 02 03 04 05 OTHERS TIME ZONE ATL DATE - REPORTS EDIT SEARCH FILES WINDOW WEEKDAY QUIT ALGORITHM 05 881001 |Saturday 03 05 | LOCAL 00 01 02 04 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 134 334 oil 164 FLL SF8 |'n 176 1U2 FLL NEI 7ts EIt 83 IF IC Lax Eve I 137 7 Lax 158 i Il: Ia .34 MC8 $F8 Po 115 117 194 799 195 83 Si8 its Ila 449 97 so 8lt8 33 12 Si8 MIto SF 88o i3 06 07 08 09 10 11 12 13 14 15 16 17 SG 28 MCI 26L 1 Eve IF Sig 146 177 LAI 8R8 IC ElR 8s 886 PB 761 Lil 187 1 SiL 18 MC8 I 8 I 57 s8 TrP 116 156 MCI EC 261 IN 1 1 56 141 iSE )5 111 161' LGA 12 116 E8 F8 19 20 21 22 23 24 01 02 03 04 05 HMT OI REPORTS OTHERS TIME ZONE ATL DATE : 881001 - EDIT SEARCH FILES II QUIT WINDOW m WEEKDAY IIIIIIIIIIN 05 |Saturdag | LOCAL 134 LAX 158 lee 3? Sig J4L SF6 97 lit Ice ILI 329 EVE -4-41 116 SF3 146 96 tax Sig 4-l LA: 117 24 ALGORITHM 116 OTHERS EDIT SEARCH REPORTS FILES WINDOW 1 WEEKDAY 1 QUIT L10 394 L 10 MCO 0240 0434 FLL 0609 0030 0454 MIA 0634 0526 0030 U 112 110 ATL 0030 0746 MCO 0902 0030 194 L10 LGA 0030 0730 PBI 0910 0340 0030 0030 83 110 839 L10 98 ATL 0100 0749 DFW 1045 Oils 0915 MIA 1058 0052 SFO 1137 0030 0431 L10 702 SFO 0520 0605 ATL 137 L10 0604 MCO 0715 0030 ATL 197 L10 0639 MIA 0815 0100 ATL 449 D L10 0932 LAM 1101 0139 ATL 117 0 110 1023 EWR 1215 0120 FLL 194 L10 1027 MCO 1140 0110 -DFW 195 L10 Ps1 1320 799 D L10 1157 SJU 1600 0105 ATL 83 D L10 0115 1200 ORD 1309 0427 ATL 83 D L10 0030 1207 MCO 1320 0105 ATL 155 D TIME ZONE 881001 D L10 BOS 0035 1155 PBI 1422 D L10 LAX 0112 1655 ATL 1865 :- L10 05 0400 PBI 0700 0030 ATL 494 D LO 1530 ATL 1750 0107 LAX 167 D LO 0108 0108 158 0326 1150 0052 KI TES -05 TIME ZONE :- 881001 364 ALGORITHM LO REPORTS I[-- ]E* I STATION ACTIVITY ROTAT ION CHAR 06 Monday ROTATION SUMMARY FLIGHT 7 R I AUDIT CHANGES LID SCHEDULE COST 61 : us DFH MIN GROUND TIME VIOLATION 003 10 SIFa L08 44A TL PBI LGAs LAX AT137:CO:1IZATL 26 0U8W11 ' 010 01 LIs JFa 12 FLL :364 ATL 44 FLL MIA a L13 ool FLL t7s 4 LAX as 56 TPA L16808 ==== um ENR a14 a85 a a aTm _= 2W~2I 14 NCO ATL: t2 BOS se n:FLL 2Al: FLL LGA 76 s26 : ER DF D13 se NCO s ATL: ATL is aaATL Lo15 g 108 s: PI ATL : JFK is 1 g 14: FLL a2aa EHR 6s 200 1. :IDA: 62 BOSs1L l F :6s: EMR : 32507;? ATL is1 FLL tt a is% :LGA: FLL J44 at lOS: 14 146: 134' AL sw ORD a LAX = 0: ILI ATL 261 261 DFN ATL si ssIA ATL :t$ ATL is: OA: LI10 LI0 se is ATL :2 SJU SWITCH RESWITCH ER 14i LAX ' SEA LAX: n MCO tit AU. tui BOS 676 IRREGULAR BLOCKTIME 261 NCO : LAX s2 :: DFH 1 a DFN its ATL: sl ss5 NCO EURDATL LAYOVERS NEEDED NCO J s MARKET SUMMARY L02 3 glo 057 ATL: I S1 3 SCHEDULE aAE LI0D ALGORITHM QUIT WEEKDAY WINDOW FILES SEARCH UTILIZATION JFK: NCO is@aNCO lzt 1s2 FLL ATL ts6 2 SEARCH REPORTS OTHERS 8 so 87 -e Monday delta-new ROTATION CHART 05 e9 16 1 2___3 1,4 15is ORD ENR: L10 17 i 10 ATL : | | UVaWe 89ITT3U A FLL I 364 ATL :44 a :FLL (1iMCO 14 SP 2 ATL: 14 PRI pesmm 1m ae40 is FLL ATL: 112 NCO imm 1 woK; SFO O ENR 126 1 134 ENR :ENR: 005 t 1 msao | FLL 6es sammu - LGA Ipew too Os FN NCO 2 :BDA: s MCO se A: | FLL 10 isi JFK 144 n LGA: is 04.0 iu FLL BOS se ATL :46: i34 gsm m FLL t6 B6S i ATL 10 Lo6 ulgg 14 21 FLL :176 ATL us TPA fmm 1eewt vi s :ATL iss: 1m UiIP87 t4FM9 MIA 13-107~~~ 1 :T ORD I L12g 68168 L15008 15 s TL LAX u ist JFK I AT DFN: sw 71 1m I 20 s: :S53I LoI ATL 2s n61 le,7 I ATL ; a00016 g I; is s7 LAX 10 : SEA '1 ATL m37:MCO : 1ATL %s 6MIA es :LAX 70 : ENR i LAX ATL BDA: s1 BOS: 4S4 ATL PBI :1s4 LGA LAX t£1 e,4 PRI: 12 :LAX 15Is7a as: :2' pee NCO: 93 u6 7f5 pt 1A54 02 DFN uS 1:-24 SJU ss ATL el ATL ol NCO tit ATL :DFN: 1oepee 3 ATL: 73 MIA: 3s NCOsk 24 :m BOS in TPA 105 MCO is see5 ,eye 23 : ss6 : DFN 676: 1 22 : SJU : 1m eis PBI 0040@s NCO 21 32 3'2pi : 056 s 2,0 t :LAX iie 1,9 ATL tss 43 SEA: is ATL: 1 so SFO 007 ppgisameam SFO: 2088 008 LOIH QI WEDY WINDOW ES ROTATION ._. .... .. GMT UME EDI' FLGH 62' FLL s ENR :82' BOSsa BL JFK: 6s ATL : i t6W68 12 gam NCO t2 |FLL 1 M:t "p. ATL I'S6 "PKA: 29g" , , ,, 1"'1" "Alfill 01 01N hi 'Wil 1, , I1111 111W MINIMUM 11000MINNOW1114, -221- VI. STAGE 2 - AUTOMATED DECISION SUPPORT INTRODUCTION OF AUTOMATED ALGORITHMS, EXPERT SYSTEMS - to assist human schedulers with certain sub-problems - to eliminate manual effort at certain steps of process - to broaden search for optimal or good solutions to scheduling sub-problems - may introduce mainframe, large scale optimization algorithms EXAMPLES OF EXISTING AUTOMATED DECISION SUPPORT ALGORITHMS a) Best cancellation of flights given breakdowns and spares b) Least revenue loss when reducing available fleet c) Optimal switching of flights between types of aircraft d) Automatic switching for transition to new schedule plan e) Automatic weekend schedule cancellations f) Automatic holiday period rescheduling g) Minimum fleet size for given services with time windows h) Automatic gate assignment at all stations i) Automatic aircraft rotation generation (with maintenance constraints) -222- VI. AUTOMATED DECISION SUPPORT a) b) Best cancellation of flights given breakdowns and spares Least revenue loss when reducing available fleet Fleet Routing Models - use network flow algorithms OBJECTIVE: Maximize Operating Income GIVEN: - Set of potential services to be flown with fixed operating times and known net operating income Daily ownership costs of aircraft Desired overnights Fixed number of available aircraft OUTPUT: - "Best" services to be flown - Marginal value for services not flown - Marginal value of adding an aircraft to fleet WEAKNESS: - Fixed service times - Fixed net income for services i.e.no spill if not flown - Single type of aircraft-solved sequentially AMIR -223- SCHEDULE MAP STATION STATION (A STATION 6 C0600 A0600 Time Cycle or Overnight Arc, C i j= Ground arc, G lij 0 = U- -= Service Arc, S I Jj = = c- = 0 A0559 c. . 'I = 00 ci - = daily 0 rental cost of aircraft o value of service C0559 4I i LI L ROTATION 12 eMT In LAX: a a a a a delta_jvctst6 19 1 17 6 CHART is a a a a a 14 a a a a a 15 ORD: 6~1 tre ses~ ; FII ORD e?56 DFN inz SF0 a a a a a a a a a a a a a a a g a a a a a a a a a a a a a a a a a a aa a aa a a a a a a i a a 002 00100 a 0-0100 a a a a a 012 818 013008 01500100 g a a a a a a a a 02 03 ORD a a a a a a a a a a a a a a a 1~7 j~~a a a e4 0 U5 , L~X AlL : ss ORD t72 SF0: 22j2~1/ a g a 01 22J231 a a 24 RDa 1 a a 23 :ORDU 'i;igu 006 016 0--011881 00800100 22 ssa :i g a 007 21 A TL a 005 ALGORITHM QUIT Monday 20 : :si WEEKDAY WINDOW FILES SEARCH EDIT REPORTS OTHERS a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a - - CMTN IV ) a 1I 12 al 1n I9 7_10 L 20 21 22 L00B ORD: s~ LAX: DFN i SF0 92 al :OR[D I II II II 7i II *I II *I *I 007ole * I I I I I I 408 I * 00901oo I 007 00100j 0080100 I I I I I I I I I I I 013011 I I I * 012011 I I I I I * 0 I I I I I I I I I I I I ~'~ I I I I I I I g I I I I I I I I I I :ss I I I I I I I I I I I Ig I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I *I I I I I *I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I 014661001 I I I I 016ggi00 as I I I I I I I I I ORD 8 8 07 I I I I I I I I I * I I * * I I I I i7 -7217 , I I I I I * I I I I I I I I I * I I I I I - I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I * I * - I I I I *I *I *I I I I I I I I I II I II II II I *I II 95 I * I g I I I g Trjq 11 1 I I I I I * * 94 * *I * I I I I I I I I I ATL *I II oI *I *I II 16 Ga : ei LAX OR :1s:? :sn 24 ORI :.::ORD:ses:DFH aim :ATL: se 00 23 I I I I I I I I I I I I I * * * I I I I I I I I I * I I I I * I * * * I I I I I I I I I I I I I I I I I I I I I * I * I I I * I * I * I I * I I I I I I I I I I I I I I I I I I I I I I I I I I I I I * I * * I I * * I I I I * I * I * * I I I I * I * * I I I * I I I * I * * * I I I I I I * I I I I I I I I I I * I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I S I I I I I I I I I I I I I I I I I I I I I I I I I I I -226- VII. SUMMARY - STATE OF THE ART IN COMPUTERIZED SCHEDULING Conclusions 1. We cannot create analytical models which are adequate to describe mathematically the complete airline scheduling problem. 2. For existing models which promise utility, we generally do not have the correct data inputs, and it isdifficult to conceive of creating the necessary models for passenger behavior in today's competitive markets. The existence of large scale solution techniques is not sufficient to justify their use at present. 3. We can provide quick, accurate answers to many subproblems which occur in the complete scheduling process, but we need an environment which allows these techniques to be available to human schedulers. This environment isnow available in the form of a network of computer workstations. 4. It isattractive to consider a single, integrated system to be used by various airline personnel as the scheduling process moves from initial planning to final execution. 5. People will remain an important part of the airline scheduling process. They are responsible for generating good schedules, and need "decision support" in their activities. There never will be a "push - button" scheduling system. 6. The desired approach isan incremental introduction of computerized assistance via graphic workstations. The strategy should be to create evolutionary stages: Stage 1 - Introduce the Scheduling Workstations Stage 2 - Introduce Automated Decision Support Stage 3 - Extend to real time Execution Rescheduling AN, I 111HINIIIIII III III ii I, INN -227- VI. SUMMARY - STATE OF THE ART IN COMPUTERIZED SCHEDULING (con't) 7. The scheduling process is not permanent - as time goes by, the problems change (perhaps temporarily), and the markets evolve, and there will be emphasis on different aspects. It will not be possible to create a completely automated decision maker which keeps up with changes. 8. As these tools are developed, they have their impact on the Scheduling Process - it will change in its flow of information, the sequence of processing will change, and eventually the airline's organizational structures will change. The introduction of computer automation must be adaptive to allow these changes to occur. 9. Every airline will have to develop its own automated scheduling system and manage the evolutionary impact on its operations. There isno single, turnkey solution to be provided by outsiders. A conceptual, long term plan is needed to direct the evolutionary effort and prevent building an incoherent set of subsystems.