Air Transportation Network Load Balancing using Auction-Based Slot Allocation for Congestion Management

Transcription

1 Air Transportation Network Load Balancing using Auction-Based Slot Allocation for Congestion Management GL Donohue, L Le, CH Chen, D Wang Center for Air Transportation Systems Research Dept. of Systems Engineering & Operations Research George Mason University Fairfax, VA NEXTOR Wye River Conf June 21-23, 23, 2004

2 Outline Description of the Congestion Problem Chicago O Hare Airport NY La Guardia Airport History of Congestion Management in the US Auction model for airport arrival slots Chicago ORD airport case study simulated scenarios results and interpretation Observations and Recommendations

3 National Airspace System Characteristics The NAS is a Stochastic Adaptive Network Stochastic: The system is characterized by PDF s Adaptive: These PDF s are a function of the System State and Airline Market Decisions Reasons that the NAS Cannot be Deterministic: Weather (winds, hazardous weather) Mechanical Equipment Characteristics Air Traffic Control System (including Controllers) Aircraft Control System (including Pilots) Airline Schedules set by varying Market Conditions All Analysis and FAA Rules Must Acknowledge this Fundamental Nature in the Future

4 Operations are Back but ORD Delays are Worse Monthly Total Operations at Major US Airports (Source: ETMS) 90,000 85,000 80,000 75,000 70,000 65,000 60,000 Jan-01 Mar-01 May-01 Jul-01 Sep-01 Nov-01 Jan-02 Mar-02 May-02 Jul-02 Sep-02 Nov-02 Jan-03 Mar-03 May-03 Jul-03 Sep-03 Nov-03 Jan-04 Air travel is gradually picking up #Flights ATL ORD Month Delayed Flights of Major US Air Carriers at ORD (Source: BTS) 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Jan-01 Mar-01 May-01 Jul-01 Sep-01 Nov-01 Jan-02 Mar-02 May-02 Jul-02 Sep-02 Nov-02 Jan-03 Mar-03 May-03 Jul-03 Sep-03 Nov-03 Jan-04 Congestion is coming back #Flights Departure Arrival Month

5 Chicago O Hare: A Maximum Capacity Hub Airport Runway layout: Departure/Arrival Pareto Trade-off: FAA/DoT 2001 Benchmark Report ASPM Data April 2000 VMC IMC Capacity

6 ORD in Dec 2003: Result of HDR Removal ASPM Data Dec 2003 DoT/FAA 2001 Benchmark Report: Data April 2000 VMC

7 Scheduled Arrival Rate for Different Airlines at ORD in Nov. and Dec Scheduled Arrival Rate (15 min time block) # of scheduled arrivals Time (15 min time block) The Green line = 21 arr. /15 min, upper-bound of IMC AR at ORD in Nov. and Dec The Red line = 25 arr./15 min, Max AR from FAA Benchmark Report for all arrivals Data from BTS which only includes domestic flight data for 15 certificated airlines

8 Smaller Aircraft trend Exacerbates Congestion Frequency competition Reduces Seat Capacity and Increases FAA Operational Load O Hare Airport Yearly Throughput 72,501,988 72,610,121 72,145, , , , ,917 69,508, , ,228 67,448,064 66,565, Year Flight Operations Passengers

9 Enplanement Capacity vs. Operational Capacity Small aircraft make inefficient use of runway capacity: 50% Flights Provide 70% Enplanement Opportunities ORD Scheduled Operations (BTS Dec 2003) cumulative seat share seats/aircraft cumulative flight share

10 NY LaGuardia: A Maximum Capacity non-hub Airport 1 Arrival Runway 1 Departure Runway 45 Arrivals/Hr (Max) 80 Seconds Between Arrivals 11.3 minute Average Delay 77 Delays/1000 Operations 40 min./delay TOTAL SCHEDULED OPERATIONS AND CURRENT OPTIM UM RATE BOUNDARIES Schedule Facility Est. Model Est.

11 NY LaGuardia :A Maximum Capacity Non-Hub Airport Departure/Arrival Pareto Trade-off: Runway layout: ASPM Data April 2000 VMC ASPM - Apr Visual Approaches ASPM - Oct Visual Approaches Calculated VMC Capacity Optimum Rate (LGA) Arrivals per Hour ,40 Each dot represents one hour of actual traffic during April or October Departures per Hour IMC Capacity

12 Factors that Determine Capacity Local Airport Authority #Runways, #Taxiways, High-speed turnoffs, #Gates, RW spacing, RW configuration, Noise Restrictions, etc. FAA ATM/CNS Equipage, Separation Standards, ATC Procedures and Airspace Design Weather Winds, Ceiling, Visibility, Severe weather Airline Schedules Network Banking Requirements Market Competition Strategies

13 History of US Slot Management LGA Airport Slot Control High-Density Rule Slot ownership AIR-21 Lottery End of HDR. What s next? 1968 Deregulation Apr 2000 Jan Limited #IFR slots during specific time periods -Negotiation-based allocation -ORD, LGA, DCA, and JFK Use-it-or-loseit rule based on 80% usage Exempted from HDR certain flights to address competition and small market access Jun 2002 HDR removed Jan 2004 June 2004 FAA-regulated 5% reduction -Congestion pricing? -Auction? FAA-regulated 2.5% reduction ORD Airport Slot Control

14 DOT s congestion management LGA Airport Slot Control AIR-21 Apr 2000 Exempted from HDR certain flights to address competition and small market access Nov 2000 Over- Scheduling Slottery Jan 2001 Sep 2001 Major chokepoint 300 new daily flights 25% total network delay Oct 2001 Jun 2002 HDR removed 130 more flights 42% delayed (from 18% in 2002) 13 mishaps (from 3 in 2002) Over- Scheduling Nov 2003 Jan 2004 June 2004 FAA-ordered 5% reduction by AA and UA ORD Airport Slot Control 2007 End of HDR FAA-regulated 2.5% reduction by AA and UA

15 Congestion Management Approaches Administrative negotiation-based IATA biannual conferences Market Based weight-based landing fee: no incentive for large aircraft inefficient Enplanement capacity time-based congestion pricing: not reveal the true value of scarce resources DoT/FAA supervised Market-based Auctions of Arrival Metering-Fix Time Slots Hybrid

16 Auction Model Design Issues Feasibility Package slot allocation for arrival slots Politically acceptable net prices Optimality Efficiency: i.e. Match Customer value to Cost Maximum Schedule Predictability Optimum airline schedule and aircraft assignment Minimum passenger ticket price Regulatory standards: capacity, international flight priorities Equity: Stability in schedule Airlines need to leverage Prior Investments Airlines competitiveness : new-entrants vs. incumbents Flexibility Primary market at strategic level Secondary market at tactical level

17 Design Approach Objective: Obtain Better Utilization of Nation s Airport Network Infrastructure Network Load Balancing Provide Cities an Optimum Fleet Mix Ensure Fair Market Access Opportunity Increase Schedule Predictability - reduced queuing delays Assumptions Airlines will make optimum use of slots they license Auction rules: Bidders Could Be Ranked using a linear combination of: Monetary offer (combination of A/C equipage credit and cash) Flight OD pair (e.g. international agreements, etc.) Airline s prior investment? On-time performance?

18 Strategic Auction Analytical Approach Auction Model Network Model Bids Schedules Airlines Slots Auctioneers NAS -Auctions only at Capacitated Airports -Auction Licenses good for 5 to 10 years Analysis & Feedback

19 Network Model used to Evaluate Auction Effectiveness 13-node network SFO LAX Runway capacity determined by Wake Vortex Separation Standards (nmiles/seconds) (M. Hanson) PHX Trailing aircraft Small Large B757 Heavy Leading aircraft Small 2.5/80 2.5/68 2.5/66 2.5/64 Large 4/ /73 2.5/66 2.5/64 B757 5/201 4/115 4/102 4/101 Heavy 6/239 5/148 5/136 4/104 and a scale factor to account for runway dependency DEN MSP ORDDTW DFW ATL LGA IAD BWI departure separation arrival separation

20 Auction Model Process Set initial price for each 15-min interval An adaptation of Clock Combinatorial Auction Model (Porter, D., Rassenti, S., Roopnarine, A., and Smith, V.) No Call for demands More demand than capacity Yes Raise price clock by a fixed increment For each auctioned interval Submit slot demands and flight info. Solve combinatorial package bidding Yes Clock price raised? No End auction process Auctioneer s action Airline s action

21 ORD Simulated Arrivals in VMC No Auction Domestic Arrivals (Dec 2003) #Flights Scheduled (BTS) simulated mean Mean of Estimated Average Arrival Delay (good weather) Min Time

22 ORD Simulated Schedule and Delay: 3 Auctioned Capacity Limits (VMC-IMC) Scheduled Domestic Arrivals at ORD Min Ar/15 min 18 Ar/15 min Time Before Estimated auctionrunway After Queuing auction Delay auctioned at 18ops 15 Min Time Before auction After auction auctioned at 18ops

23 De-peaking results in Significantly Improved Passenger Schedule Predictability Average Estimated Flight Delay ASPM - April Instrument Approaches ASPM - April Visual Approaches 400 Arrivals per Hour Calculated VMC Capacity Calculated IMC Capacity Optimum Rate (ORD) 100,100 80,80 Min IMC Current Schedule auctioned at 21 ops/15 min auctioned at 18 ops/15 min VMC Departures per Hour Quarterly Airport Arrival Rate in Different Weather Conditions

24 Auction Produced Rolling Banks Changes the Distribution of Delays Estimated Arrival Queuing Delay 0.25 Current Schedule Auction-produce schedule Probability of Delay Duration min

25 ORD Flights to be Rerouted vs. Average Actual Cancellations / day Estimated #flights to be rerouted #Flights ops/15 min 21 ops/min Current Schedule Quarterly Airport Arrival Capacity in Different Weather Conditions Daily number of cancellations (BTS)

26 Significantly Improved ORD Passenger Schedule Predictability Auction Produced a Coordinated Airline De-Peaked Schedule Simulated ORD Auction at 21 Arrivals/15 Min: Reduced Delays by Over 80% Required only 26 Flights to be Re-Scheduled through another Non-Capacitated Hub Airport This Reduction is Comparable to the Reported Daily Flight Cancellation Rate

27 Research Issues to be Addressed Who is Eligible to Bid? Airlines, Airports, General Aviation, Investors What is the Fundamental Bidding Metric? $/15 min 95% Confidence, $-Passenger/Aircraft Slot How Many Slots Should be Auctioned Delay (min))? VMC N(4,2 2 ), IMC N(8,4 2 ), IMC N(15,8 2 ) What Bid Combinations will be Allowed? Packages w/ Ranked Priorities, Intercity Packages, etc. Bidding Activity Rules What are the Payment Options? Up Front for X yr. Lic., Monthly Royalty Payments for X Yrs. Who gets the Money? Airports (PFC Sub.), Airlines (Equip. Vouchers), FAA (Ticket Tax Sub.) What are the Secondary Market Rights? What is the Winner Determination Algorithm? Auction Frequency/Duration of Slot License? License for 5 yr., 10 yr.,?

28 Observations on Research to Date Combinatorial Clock Auctions Offer a Promising Market-Based approach to Congestion Management Auction Proceeds could be used as Incentives to the Airports for Infrastructure Investments and to the Airlines for Avionics Investments Congestion Management at Critical Network Node Airports will have a Profound Effect on Increasing Passenger Travel Predictability Simple Auctions might Exclude Small Aircraft and/or Small markets from Hub Airports Simple Bidding Rules can Prevent this Problem

29 Future work Conduct 3 FAA Strategic Simulations to Resolve Slot Allocation Issues First Simulation would Examine a Variety of Policy Problems/Options (Include a broad collection of Stakeholders) Second Simulation would examine specific sets of auction rules and instruments Third Simulation would use Results of first two to Evaluate Modified Congestion Mgt. Options Continue Model development to Refine Combinatorial Package Bidding Simulations to Evaluate Proposed Auction Rule Set

30 Backup

31 Simulation Model for Testing Auction Design General Assumptions: Aircraft can arrive within allocated 15 min. Arrival Time slots with Required Time-of-Arrival errors of 20 seconds (using Aircraft RTA Capabilities) Auction items: Metering Fix Arrival Slots in min bins (5:00am till 24:00am) up to 21 arrivals/bin Input: Dec 2003 BTS schedule of 2186 flights domestic flights to ORD (80% of total traffic) ORD Scheduled Arrivals (Source: ASPM, BTS Dec 2003) Total Arrivals Domestic Arrivals of 15 certificated airlines 30 #Ops Time

32 Airline model assumptions Single market, single item bidding Airlines flexibility for changing schedule: one 15-min bin 15min 15min u(b i ) = 1 Linear decreasing bids withdrawn original scheduled 15-min interval bids withdrawn 0 b i b i+1 b i-1 Homogenous and honest bidding with upper threshold proportional to aircraft size

33 Airline Package Bidding Model Target Slots P j : Possible packages P jk for P j : b i lb b i b i ub Δb i,i+1 lb b i+1 b i Δb i,i+1 ub LP Model: Maximise Subject to: k k x 1 Π j k j j k P k u ( P j ) x k j x j k j b 1 b 2 k j B Utility Function u(p j1 ) = 1 non flexible schedule u(p jk ) = 0 P 1 j P 2 j highly flexible schedule P j k Variables: {P jk } set of package bids B airline bidding budget k 1 if airline bids for package P k x j = j 0 otherwise Π price vector

34 Network Simulation Model used to Evaluate Auction Effectiveness Arrivals per Hour ASPM - April Instrument Approaches ASPM - April Visual Approaches Calculated VMC Capacity Calculated IMC Capacity Optimum Rate (ORD) 100,100 80,80 Stochastic Queuing Model 12 Capacitated Airports 1 Airport Unconstrained sink and source ORD Runway capacity determined by Wake Vortex Separation Standards (nmiles/seconds) (M. Hanson) Trailing aircraft Small Large B757 Heavy Leading aircraft Small 2.5/80 2.5/68 2.5/66 2.5/64 Large 4/ /73 2.5/66 2.5/64 B757 5/201 4/115 4/102 4/101 Heavy 6/239 5/148 5/136 4/104 and a scale factor to account for runway dependency weather effect Departures per Hour Delay = Arrival Delay + Queuing Delay Good weather Condition: N(0,5 2 )

35 UP-Front Payment vs. Cash-Flow Royalty Auction Proceeds could be paid out to the FAA on a monthly basis (i.e. License Royalty Fee to Reserve Arrival Time Slot) FAA could retain a % to replace ATC ticket tax Airport could use a % to replace PFC tax and invest in New Runways, Taxiways, etc.

36 Airline Avionics Investments Required to Increase Airport Capacity Flight Management Systems with Required Time of Arrival Capabilities ADS-B Cockpit Display of Traffic Information with the Capability of Providing Pilot Controlled Time-Based Separation

37 Airlines Could bid with Avionics Investment Promissory Notes Airlines could Bid with Script that constituted a contract to equip their Aircraft with-in X years (i.e. ½ bid price) Accepted Airline Bid constitutes a Contract with the FAA to provide Operational Procedures that Utilize Decreased Separation Capabilities

38 LGA Arrival - Departure IMC ASPM - April Instrument Approaches ASPM - October Instrument Approaches Calculated IMC Capacity Reduced Rate (LGA) Arrivals per Hour , Departures per Hour

39 Dynamics of Over-Scheduling Flight banking creates inefficient runway utilization ORD Scheduled Operations (Source: ASPM Dec 2003) Total Operations Arrivals optimum total rate Airline competition for market share UA AA and UA Scheduled Operations at ORD grouped by 15-min epochs (Source: BTS Dec 2003) Departure Arrival Time optimum arrival rate AA