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, 2004
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
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:
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
Operations are Back but
ORD Delays are Worse
Monthly Total Operations at Major US Airports (Source: ETMS)
90,000
ATL
85,000
Air travel is
gradually
picking up
ORD
80,000
#Flights 75,000
70,000
65,000
M
Ja
n
-0
1
ar
-0
M 1
ay
-0
Ju 1
l-0
Se 1
p0
N 1
ov
-0
Ja 1
nM 02
ar
-0
M 2
ay
-0
2
Ju
l-0
Se 2
pN 02
ov
-0
Ja 2
n0
M 3
ar
-0
M 3
ay
-0
Ju 3
l-0
Se 3
p0
N 3
ov
-0
Ja 3
n04
60,000
Month
Delayed Flights of Major US Air Carriers at ORD (Source: BTS)
Arrival
n0
M 1
ar
-0
M 1
ay
-0
1
Ju
lSe 01
p0
N 1
ov
-0
Ja 1
n0
M 2
ar
-0
M 2
ay
-0
2
Ju
l- 0
Se 2
p0
N 2
ov
-0
Ja 2
n0
M 3
ar
-0
M 3
ay
-0
Ju 3
lSe 03
p0
N 3
ov
-0
Ja 3
n04
Departure
Ja
Congestion is
coming back
14,000
12,000
10,000
8,000
#Flights
6,000
4,000
2,000
0
Month
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
ORD in Dec 2003:
Result of HDR Removal
ASPM Data Dec 2003
DoT/FAA 2001 Benchmark Report:
Data April 2000 VMC
Scheduled Arrival Rate for Different
Airlines at ORD in Nov. and Dec. 2003
# of scheduled arrivals
Scheduled Arrival Rate (15 min time block)
Time (15 min time block)
•
The Green line = 21 arr. /15 min, upper-bound of IMC AR at ORD in Nov. and Dec. 2003
•
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
Smaller Aircraft trend
Exacerbates Congestion
ƒFrequency competition Reduces Seat Capacity and
Increases FAA Operational Load
O’Hare Airport Yearly Throughput
72,501,988
72,610,121
922,817
911,917
908,989
896,104
928,691
72,145,489
69,508,672
67,448,064
896,228
66,565,952
1998
1999
2000
2001
2002
Year
Flight Operations
Passengers
2003
Enplanement Capacity vs.
Operational Capacity
ƒSmall aircraft make inefficient use of runway capacity:
50% Flights Provide 70% Enplanement Opportunities
ORD Scheduled Operations (BTS Dec 2003)
cumulative
seat share
1
500
0.9
450
0.8
400
0.7
350
0.6
300
0.5
250
0.4
200
0.3
150
0.2
100
0.1
50
0
0
0.1
0.2
0.3
0.4
0.5
0.6
cumulative flight share
0.7
0.8
0.9
1
seats/aircraft
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
T O T A L S C HE D U LE D O P E R A T I O N S A N D C U R R E N T O P T I M U M
R A T E B OU N D A R IES
32
28
24
20
16
12
8
4
0
7
8
9
10
11
12
13
Sched ule
14
15
Facilit y Est .
16
17
M o d el Est .
18
19
20
21
NY LaGuardia :A Maximum
Capacity Non-Hub Airport
Departure/Arrival Pareto Trade-off:
ASPM Data April 2000 VMC
Runway layout:
ASPM - Apr 2000 - Visual Approaches
60
ASPM - Oct 2000 - Visual Approaches
Calculated VMC Capacity
50
Arrivals per Hour
Optimum Rate (LGA)
40
40,40
Each dot represents one
hour of actual traffic
during April or October
2000
30
20
10
0
0
10
20
30
40
50
60
Departures per Hour
IMC
Capacity
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
History of US Slot Management
LGA Airport Slot Control
Slot ownership
High-Density
Rule
Deregulation
1968
-Limited #IFR slots
during specific time
periods
-Negotiation-based
allocation
-ORD, LGA, DCA,
and JFK
1978
1985
AIR-21
Apr
2000
Use-it-or-lose- Exempted from
it rule based
HDR certain
on 80% usage
flights to
address
competition and
small market
access
Lottery
End of HDR.
What’s next?
Jan
2001
2007
Jun
2002
Jan June
2004 2004 -Congestion
pricing?
-Auction?
FAA-regulated
2.5% reduction
HDR
removed
FAA-regulated
5% reduction
ORD Airport Slot Control
DOT’s congestion management
LGA Airport Slot Control
AIR-21
Apr
2000
“Slottery”
Nov
2000
Jan
2001
130 more flights
42% delayed (from 18% in 2002)
13 mishaps (from 3 in 2002)
Sep Oct
2001 2001
OverScheduling
Exempted from
HDR certain
flights to
address
competition
and small
market access
Major chokepoint
300 new daily flights
25% total network delay
OverScheduling
Jun
2002
Nov
2003
Jan
2004
June
2004
2007
End of HDR
FAA-regulated 2.5%
reduction by AA and UA
HDR
removed
FAA-ordered 5%
reduction by AA and UA
ORD Airport Slot Control
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
Auction Model
Design Issues
‰ Feasibility
− Package slot allocation for arrival slots
− Politically acceptable net prices
‰ Optimality
− Efficiency: i.e. Match Customer value to Cost
ƒ Maximum Schedule Predictability
ƒ Optimum airline schedule and aircraft assignment
ƒ Minimum passenger ticket price
− Regulatory standards: capacity, international flight priorities
− Equity:
ƒ Stability in schedule
ƒ Airlines’ need to leverage Prior Investments
ƒ Airlines’ competitiveness : new-entrants vs. incumbents
‰ Flexibility
− Primary market at strategic level
− Secondary market at tactical level
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 ?
Strategic Auction
Analytical Approach
Auction Model
Network Model
Bids
Schedules
Slots
Airlines
-Auctions only at
Capacitated Airports
-Auction Licenses
good for 5 to 10 years
Auctioneers
NAS
Analysis &
Feedback
Network Model used to
Evaluate Auction Effectiveness
13-node network
DEN
SFO
LAX
PHX
MSP
ORD DTW LGA
BWI
IAD
DFW
ATL
Runway capacity determined by
♦ Wake Vortex Separation Standards
(nmiles/seconds) (M. Hanson)
Trailing aircraft
Leading aircraft
Small
Large
B757
Heavy
Small
Large
B757
Heavy
2.5/80
4/164
5/201
6/239
2.5/68
2.5/73
4/115
5/148
2.5/66
2.5/66
4/102
5/136
2.5/64
2.5/64
4/101
4/104
♦ and a scale factor to account for runway
dependency
departure separation
arrival separation
Auction Model Process
An adaptation of Clock Combinatorial Auction Model
(Porter, D., Rassenti, S., Roopnarine, A., and Smith, V.)
Set initial price for
each 15-min
interval
Submit slot
Call for
demands and flight
demands
info.
For each
auctioned interval
Solve
combinatorial
package bidding
No
More demand
than capacity
Yes
Raise price clock by
a fixed increment
Yes
Clock price
raised?
No
End auction
process
Auctioneer’s action
Airline’s action
ORD Simulated Arrivals in
VMC – No Auction
#Flights
40
Scheduled (BTS)
simulated mean
Domestic Arrivals
30
(Dec 2003)
20
10
0
0
1
2 3
4
5
6 7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
0
1
2
4
5
6
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Min
20
Mean of Estimated 15
Average Arrival 10
Delay
5
(good weather)
0
3
7
Time
ORD Simulated Schedule and Delay:
3 Auctioned Capacity Limits (VMC-IMC)
Scheduled Domestic Arrivals at ORD
40
21 Ar/15
min
30
Min 20
18 Ar/15
min
10
0
0
3
6
9
12
15
18
21
Time
Estimated
ay Queuing
Before
auctionRunwAfter
auction Delay
auctioned at 18ops
15
10
Min
5
0
0
3
6
9
12
15
18
Time
Before auction
After auction
auctioned at 18ops
21
De-peaking results in Significantly
Improved Passenger Schedule
Predictability
Average Estimated Flight Delay
ASPM - April 2000 - Instrument Approaches
400
ASPM - April 2000 - Visual Approaches
350
Calculated VMC Capacity
160
Current Schedule
Calculated IMC Capacity
300
Optimum Rate (ORD)
140
120
Arrivals per Hour
auctioned at 21 ops/15 min
250
100,100
auctioned at 18 ops/15 min
Min 200
100
80,80
150
80
VMC
100
60
50
40
IMC
0
20
14
0
0
20
40
60
80
100
120
Departures per Hour
140
160
15
16
17
18
19
Quarterly Airport Arrival Rate
in Different Weather Conditions
20
21
Auction Produced Rolling Banks
Changes the Distribution of Delays
Estimated Arrival Queuing Delay
Current Schedule
Auction-produce schedule
0.25
Probability
of Delay
Duration
0.20
0.15
0.10
0.05
0.00
3
8
13
min
18
ORD Flights to be Rerouted vs.
Average Actual Cancellations / day
Estimated #flights to be rerouted
Daily number of cancellations (BTS)
350
300
#Flights
18 ops/15 min
250
250
21 ops/min
200
Current Schedule
200
150
150
87
100
26
50
0
50
14
15
16
17
18
19
20
Quarterly Airport Arrival Capacity
in Different Weather Conditions
73
100
26
29
29
2.2004
3.2004
21
0
12.2003
1.2004
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
Research Issues
to be Addressed
‰ Who is Eligible to Bid?
− Airlines, Airports, General Aviation, Investors
‰ What is the Fundamental Bidding Metric?
− $/15 min Slot @ 95% Confidence, $-Passenger/Aircraft Slot…
‰ How Many Slots Should be Auctioned (arr @Prob. Delay (min))?
− VMC ROT @ N(4,22), IMC WV @ N(8,42), IMC WV @ N(15,82) …
‰ 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., ?
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
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
‰Backup
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 76 15-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
50
Domestic Arrivals of 15 certificated airlines
40
30
#Ops
20
10
0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23
Time
Airline model assumptions
ƒ Single market, single item bidding
ƒ Airlines’ flexibility for changing schedule: one 15-min bin
u(bi) = 1
15min
bids
withdrawn
Linear
decreasing
15min
original scheduled
15-min interval
bids
withdrawn
0
bi
bi+1
bi-1
ƒ Homogenous and honest bidding with upper threshold
proportional to aircraft size
Airline Package
Bidding Model
Utility Function
Target Slots Pj:
b1
b2
Æ Possible packages Pjk for Pj:
bilb ≤ bi ≤ biub
Δbi,i+1lb ≤ bi+1 – bi ≤ Δbi,i+1ub
u(Pj1) = 1
highly flexible
schedule
non flexible
schedule
u(Pjk) = 0
P j1
P j2
P jk
LP Model:
k
k
u
(
P
)
⋅
x
∑∑ j j
Maximise
j
k
Variables:
Subject to:
∑x
k
j
≤1
∀j
k
∑ ∑ Π ⋅ Pjk ⋅ x kj ≤ B
j
k
{Pjk}
B
⎧1
x =⎨
⎩0
k
j
Π
set of package bids
airline bidding budget
if airline bids for package Pjk
otherwise
price vector
Network Simulation Model used to
Evaluate Auction Effectiveness
– Stochastic Queuing Model
– 12 Capacitated Airports
–1 Airport Unconstrained sink and source
– ORD Runway capacity determined by
ASPM - April 2000 - Instrument Approaches
ASPM - April 2000 - Visual Approaches
Calculated VMC Capacity
160
Calculated IMC Capacity
Optimum Rate (ORD)
140
Trailing aircraft
Leading aircraft
Small
Large
B757
Heavy
120
Arrivals per Hour
ƒ Wake Vortex Separation Standards
(nmiles/seconds) (M. Hanson)
100,100
100
80,80
80
Small
Large
B757
Heavy
2.5/80
4/164
5/201
6/239
2.5/68
2.5/73
4/115
5/148
2.5/66
2.5/66
4/102
5/136
2.5/64
2.5/64
4/101
4/104
ƒ and a scale factor to account
9for runway dependency
9weather effect
60
40
20
0
0
20
40
60
80
100
120
Departures per Hour
140
160
– Delay = Arrival Delay + Queuing Delay
Good weather Condition:
N(0,52)
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.
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
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
LGA Arrival - Departure IMC
60
ASPM - April 2000 - Instrument Approaches
ASPM - October 2000 - Instrument Approaches
Calculated IMC Capacity
50
Arrivals per Hour
Reduced Rate (LGA)
40
32,32
30
20
10
0
0
10
20
30
40
Departures per Hour
50
60
Dynamics of Over-Scheduling
Flight banking creates
inefficient runway utilization
Airline competition for
market share
AA and UA Scheduled Operations at ORD
grouped by 15-min epochs
(Source: BTS Dec 2003)
ORD Scheduled Operations (Source: ASPM Dec 2003)
80
Total Operations
20
Arrivals
70
Departure
16
60
optimum
total rate
50
40
Arrival
12
UA
30
optimum
arrival rate
20
8
4
10
0
0
5
8
11
14
Time
17
20
23
0
4
8
12
AA
16
20