Underlying Problems and Major Research Issues Facing the US Air Transportation System George L. Donohue, Ph.D. Professor, Systems Engineering and Operations Research Director, Center for Air Transportation Systems Research 2nd International Conference on Research in Air Transportation - ICRAT 2006 Belgrade, Serbia and Montenegro June 24, 2006 CENTER FOR AIR TRANSPORTATION SYSTEMS RESEARCH Credits Research Team at GMU that have contributed to these Insights: • • • • • • • • • • • Rudolph C. Haynie, Ph.D. (2002), Col. US Army Yue Xie, Ph.D. (2005) Arash Yousefi, Ph.D. (2005) Loan Le, Ph.D. Candidate (expected 2006) Danyi Wang, Ph.D. Candidate Babak Jeddi, Ph.D. Candidate Bengi Mezhepoglu, Ph.D. Candidate Dr. Lance Sherry, Exec. Dir. CATSR Dr. John Shortle, Assoc. Prof. SEOR, CATSR Dr. C.H. Chen, Assoc. Prof. SEOR, CATSR Dr. Karla Hoffman, Prof. SEOR, CATSR CATSR Outline Worldwide Generic Problems in Air Transportation Economic System of Systems Stochastic Safety Process Control Airspace Designs are not Optimum US has some Unique Problems in Air Transportation Little Concern for Passengers Quality of Service Airport Congestion Regulations Chaotic Future Research should focus more on: Passenger Metrics and less on Aircraft Operations Metrics Stochastic Metrics and Regulations Economic System Control Mechanisms CATSR CATSR Economic System of Systems Air Transportation is a Complex Adaptive System (CAS) Problem CATSR • Essential Elements of a CAS: • Complex – Multiple Agents with many variables always working on the Edge of Stability – Possess Strong Non-linear Interrelationships but Try to bring some Order out of Chaos • Spontaneous and Self Organizing – Multiple Independent Agents Optimizing different Object Functions (i.e. constantly Learning and Adapting) • Evolutionary – constantly demonstrating Emergent Behavior • Requires a Different Modeling Approach that Includes ALL Relevant Strong Feedback Loops Air Transportation System: Agents, Inter-relationships, Adaptive Behavior and Stability CATSR Capacity Offset Suppliers of Air Traffic Flow Services Suppliers of Air Traffic Infrastructure Suppliers of Air Transportation Services Total Seats Seats, Parking, Rental Cars Enplanements Demand for Air Transportation Services Regional Markets (Businesses, Citizens) ( = weeks, Variations: Daily, Weekly, Seasonal, Econ Cycles) CAS Control Problem: Example Question What is Impact of ADS-B ? Plausible Futures? ADS-B Initiatives Delays, Flat Fees & Taxes Seats, Parking, Rental Cars Airfares, + fees, taxes, delay costs Enplanements Regional Markets (Businesses, Citizens) ( = weeks, Variations: Daily, Weekly, Seasonal, Econ Cycles) CATSR Modernization requires understanding system “pressure points” and “tipping points” (i.e. nonlinearities) Signaling Mechanisms DRIVE Air Transportation System • Balance Capacity and Demand (by signaling scarce resources) • Incentivize Innovation Strong Signals (i.e. PRICES) yield: • Effective Use of Scarce Resources (e.g. yield management, aircraft assets,…etc) • Vibrant Innovation in Airlines, and Aircraft Manufacturers sectors (see Real Yield) Weak Signals (e.g. Delays, Flat Fees & Taxes) yield: • Unpredictable day-to-day Operations • Difficulty Valuing Service (e.g. Airport Landing Slots, Labor Salary Negotiations) • Dormant Innovation Cycles Dr. Lance Sherry and Benji Mezhepoglu CATSR Stochastic Safety Process Control - Solid Theoretical Foundation NOT BEING APPLIED TO ATM Air Transportation Safety is a Stochastic Characterization and Control Problem CATSR • International Safety Standards do not recognize that they are Regulating Stochastic Processes that have at least 2 Statistical Parameters that MUST BE CONTROLLED • Research results of : • • • • Dr. Rudolph C. Haynie (2002) Dr. Yue Xie, (2005) Mr. Babek Jeddi, (in progress) Prof. John Shortle Operations Around a Typical High Capacity US Airport (Mr. Babak Jeddi, research in progress) Detroit Airport (DTW) CATSR Sample Landings on 21L: GMU Processed Multilateration Data Distorted Scale Correct Scale CATSR Data Analysis Process to Estimate: IAT, IAD and ROT pdf’s Airplane i Threshold Airplane i+1 Aircraft Type Heavy Large Large Small Runway Threshold 10:23:14 10:24:28 10:26:16 10:28:32 Leave Runway 10:24:04 10:25:13 10:27:12 10:29:28 ... ... ... Col. Clint Haynie, USA PhD., 2002 Yue Xie, PhD. 2005 CATSR Runway Occupancy Time (ROT) at AAR = 40 Arr/Rw/Hr 49 seconds 40 Ar/Rw/Hr =90 seconds • 669 samples for all aircraft types, peak IMC periods • Sample mean is 49.1 sec. • Sample std. dev. is 8.1 sec. CATSR Inter-Arrival Time (IAT) SAFETY ? 40 Ar/Rw/Hr LOST CAPACITY • IMC • 3 nm pairs • 523 samples (during peak periods) • Fit: Erlang(40;11,6): mean 106 sec, std. dev. 27 sec. CATSR Inter-Arrival Distance (IAD) SAFETY ? ADS-B RSA LOST CAPACITY Schedules, TFM, RTA • IMC • 3 nm pairs • 523 samples (during peak periods) • Fit: Erlang(1.5;0.35,6): mean 3.6 nm, std. dev. 0.86 nm. CATSR ROT vs. IAT to find Simultaneous Runway Occupancy (SRO) Probability: est to be ~1 x 10-3 Runway Occupancy Time (sec) SRO Region Inter-Arrival Time (sec) • Freq (IAT < ROT) ~= 0.0016 in peak periods and 0.0007 overall (including non-peak periods) • IMC: 1 / 669= 0.0015 in peak periods • Correlation coefficient = 0.15 CATSR ROT vs. IAT to find Simultaneous Runway Occupancy (SRO) Probability: est to be ~1 x 10-3 Runway Occupancy Time (sec) SRO Region Inter-Arrival Time (sec) •Question: •Should P(SRO)= 1 x 10-6 /Arrival? 1 x 10-5 /Arrival? 1 x 10-4 /Arrival? CATSR Runway Occupancy Time (ROT) and Increased AAR to 45 Arr/Rw/HR 45 Ar/Rw/Hr • 669 samples for all aircraft types, peak IMC periods • Sample mean is 49.1 sec. • Sample std. dev. is 8.1 sec. CATSR Inter-Arrival Time (IAT) SAFETY ? 45 Ar/Rw/Hr LOST CAPACITY • IMC • 3 nm pairs • 523 samples (during peak periods) • Fit: Erlang(40;11,6): mean 106 sec, std. dev. 27 sec. CATSR CATSR New Airspace Design Paradigms ATC Workload is not Uniform and Airspace Designs are Not Optimum • Current Airspace Designs in most countries predate modern computer Modeling and Optimization era • Controller Workload can become the Capacity Limitation in some Airspace • Current Controller Workload can be Decreased with Center and Sector Optimized Re-design • All New digital Data-Link and Automation Systems will Benefit from Re-designed, workload balanced airspace Based on Research results of Arash Yousefi (2005) CATSR WL as a continuous function of Lat, Lon, and Time (Arash Yousefi, Ph.D. 2005) WLt = f( , ) where : f is a generic function t denotes the time interval CATSR Planar Projection of Workload Function ( WLt ) CATSR Results of Center Boundary Re-design: An Example CATSR CATSR Passengers are Our Forgotten Customers - They Pay the Bills & Suffer the Penalties for Poor performance Passenger Quality of Service Metrics are NOT Currently used for System Control CATSR • Most Research Emphasis has been on Flight Delay and Airline Economic Benefits from Reduced Fuel Consumption • Little attention has been placed on the Passenger Quality of Service (PQOS) or on the real Lost Human Productivity • Lost Passenger Productivity (GDP) due to System Inefficiencies may EXCEED Airline fuel burn Losses • Flight Cancellations are as Important to Understand and Model as Flight Delays Recent Observations on Flights in the US 35 OEP Airport Network (2004) • Total Passenger Trip Delay (TPTD) metric defined (Danyi Wang (2006) work in progress) • OEP 35 Airport Network: • 3,000,000 flights, 1044 segments • 20.5% delayed > 15 min (52,100,000 Hours Delayed) • 1.78% flights cancelled (34,300,000 Hours Delayed) • At $30/Hr = $2.6 Billion/yr Lost GDP Productivity CATSR The Air Transportation System can be Modeled as a Two Tiered Flow Model • A two tiered flow model: the Vehicle Tier and the Passenger Tier (Ms. Danyi Wang, research in Progress) • • • Vehicle Tier Key Performance Index (KPI): Flight Delays, # of Delayed Flights, Cancelled Flights, On-Time Flights, % of Delayed Flights, Cancelled Flights, On-Time Flights, etc. Passenger Tier KPI: Passenger Trip Delay Passenger Trip Delay = function (“Vehicle Flight Performance”, “Passenger Factor”) CATSR Strong Non-Linear Relationship Exists between Flight Disruptions, Load Factors, Time and Total Passenger Delay • Results: • Average Passenger Delay grows Exponentially with load factor, especially for days with high flight delays and cancellations. • Low Service Frequency and Flight Disruptions late in the day contribute significantly to the delay of disrupted passengers Bratu & Barnhart (2005), Bratu (2003) and Sarmadi (2004) CATSR CATSR Airports Need Some Schedule Regulation for Safe, Efficient and Predictable Transportation US Does Little to Regulate Airport Congestion • Flight Schedules Drive Much of the Flight Delays Observed in the US Air Transportation System • Schedules are Uncoordinated (Anti-Trust Laws) • Largely Unregulated by Arrival Slot Allocations • These Delays at Hub Airports Impact the entire Air Transportation Network • Regulators are Concerned about the Adverse Effects of Slot Regulation (for Congestion Management) on the Private Service Provider’s Decisions on what Markets to Serve • i.e. What network connectivity and frequency would result from profit maximizing airlines if Capacitated Airport nodes were Regulated? • This Question can be formulated as a Network Commodity Flow Optimization Problem (Ms. Loan Le, summer 2006) CATSR Excess of demand and severe congestion at NY area airports: a 40-year old reality CATSR Timeline recap of congestion management measures HDR at EWR, LGA, JFK, DCA, ORD Perimeter rule at LGA, DCA 1969 - Limited #IFR slots during specific time periods - Negotiation-based allocation early 1970s Deregulation 1978 Removal of HDR at EWR Slot ownership AIR-21 1985 4.2000 Use-it-orlose-it rule based on 80% usage Exempted from HDR at LGA certain flights to address competition and small market access Excess of demand and severe congestion at NY area airports: a 40-year old reality CATSR Timeline recap of congestion management measures Lottery AIR-21 at LGA Apr-00 Jan-01 Removal of HDR at ORD Jul-02 End of HDR. What’s next? Jan-07 Excess of demand and severe congestion at NY area airports: a 40-year old reality CATSR Timeline recap of congestion management measures Lottery AIR-21 at LGA Apr-00 Jan-01 Removal of HDR at ORD Jul-02 End of HDR. What’s next? Jan-07 Declining Trend of aircraft size: Fewer Passengers at Constant Congestion Delay CATSR Small Aircraft & Low load-factor Flights: High Delay & Lost Airline Revenue ? CATSR Congestion management options Laissez-faire: AIR-21 HDR Airport expansion Building new runway, new airport? Develop reliever airports? Administrative options: Collaborative scheduling Bilateral? Multilateral? Market-based Congestion pricing Auction Question: What is the best use of runway capacities? What markets get to stay at their current airport? What should fly to other substitutable airports? What is the right fleet mix and frequencies? CATSR Modeling airline flight scheduling: Approaches Model individual airlines – Infinite number of competition behaviors – New entrants? – Limited data and inherent data noise Model a Benevolent Single Airline – Incorporates some competition requirement – Best schedule that could be achieved benchmark for congestion management incentives – Aggregate data reduce noise Problem statement Assuming the government as a benevolent single airline in NYC, how would that airline optimize the flight schedule to LGA/EWR/JFK? CATSR New York LGA case study CATSR A few statistics: Operations Throughput: flights Average Flight Delay: Seat throughput: Average aircraft size Number of regular markets* Average segment fare: Revenue Passengers: 93,129 38 min 8,940,384 seats 96 seats 66 (277) $133 6,949,261 Modeling Assumptions target period: Q2, 2005 45 minutes turn-around time for all fleets 75% load factor Fuel cost: $2/gallons Only existing fleets Market daily frequencies and geographical distribution: actual data CATSR Results: Profit maximizing service levels CATSR for unconstrained capacity scenario (unconstrained scenario) Markets decreasing: BOS DCA FLL RDU ORD ATL PHL DFW CLT … 7446 6842 4224 3622 6248 4834 2010 2618 3224 Results: Maximizing service levels at 10 ops/runway/15min CATSR Throughput maximizing: Profit maximizing: BOS DCA FLL RDU ORD ATL PHL DFW CLT … BOS DCA FLL RDU ORD ATL PHL DFW CLT … 74 58 68 60 4444 3636 62 50 48 32 20 12 26 22 32 20 7446 6842 4424 3622 6248 4834 2010 2618 3224 Throughput Maximizing service level at 9 ops/runway/15min Throughput maximizing: BOS DCA FLL RDU ORD ATL PHL DFW CLT CHM GSO IND BUF 74 58 68 60 4444 363620 62 5044 48 3230 20 12 26 2218 32 20 26 20 18 12 18 12 22 16 CATSR Throughput Maximizing service level at 8 ops/runway/15min Throughput maximizing: BOS DCA FLL RDU ORD ATL PHL DFW CLT CHM GSO IND BUF DTW 74 58 68 60 4444 30 363620 62 5044 34 48 3230 20 12 10 26 2218 32 20 26 20 18 12 18 12 22 16 32 20 CATSR Summary of results for LGA Non-monotonic behavior for profit maximizing schedules Monotonic behavior for seat throughput maximizing schedules CATSR Directions for Future Research Future Research should focus more on: Passenger Metrics and less on Aircraft Operations Metrics Stochastic Metrics and Regulations Optimum Airport Slot Utilization Economic System Control Mechanisms Dynamic Super-Sector Designs with Optimum Convective Weather Avoidance Capability CATSR References • • • • • • • CATSR Haynie, R.C. (2002), “An Investigation of Capacity and Safety in NearTerminal Airspace for Guiding Information Technology Adoption” GMU PhD dissertation Yousefi, A. (2005), “Optimum Airspace Design with Air Traffic Controller Workload-Based Partitioning” GMU PhD disertation Xie, Y. (2005), “Quantitative Analysis of Airport Arrival Capacity and Arrival Safety Using Stochastic Methods” GMU PhD dissertation Le, L. (2006 expected), “Demand Management at Congested Airports: How Far are we from Utopia?” GMU PhD dissertation Wang, D., Sherry, L. and Donohue, G. (2006) “Passenger Trip Time Metric for Air Transportation”, The 2nd International Conference on Research in Air Transportation (ICRAT), June 2006 Jeddi, B., Shortle J. and L. Sherry, “Statistics of the Approach Process at Detroit Metropolitan Wayne County Airport”, The 2nd International Conference on Research in Air Transportation (ICRAT), June 2006 http://catsr.ite.gmu.edu/home.html