AGIFORS Annual Symposium 2005
Technical Program*

Presentations ( Based on submitted abstracts to date* )

Monday – September 26

15:00 – 16:15 – Session 1-5
Modeling Itinerary Shares with GEV Models
Greg Coldren, Anna Valicek Medal Winner 

Tuesday – September 27

08:30 – 09:15 – Session 2-1
Dynamic Capacity in Airline Revenue Management
Michael Frank, Lufthansa, Germany 

An event-driven simulation model mirroring a real world’s revenue management system is presented. Its primary purpose is to evaluate revenue impact of a continously adjusted fleet assignment during the booking period. Simulation setup, computational results and future directions are discussed.

09:15 – 10:00 – Session 2-2
Quantifying the Risks Associated with Market Performance
Tim L. Jacobs, Delta Airlines, USA 

This paper presents a probabilistic approach for quantifying the likely market performance for an airline.  This approach uses basic concepts of engineering reliability and risk to estimate the probability that revenue performance of a service exceeds costs.  This methodology also provides a mechanism for evaluating performance requirements in terms of revenues or costs for a desired risk level.  Using First‑Order Second‑Moment methods, the approach estimates the probability that revenues will exceed costs when both revenue and cost factors behave as random variables.  Conversely, this same methodology can be used to estimate the necessary revenue performance metrics to meet a prescribed probability.  This paper will present several basic examples of this methodology.

10:30 – 11:15 – Session 2-3
Facing weather disruptions with Global Rescheduling
Cyrille Gueguen, Air France, France

When a major disruption occurs in an airport due to weather conditions or strikes, the OCC or Scheduling departement have to evaluate as fast as possible the impacts of the disruption in order to decide how to return back to schedule. Integrated in the SIROCCO real time system, the goal of the GRS tool is to provide a set of flights to cancel or delay and other flight type modifications. The given solution is driven by user strategy and respects aircraft, crew and passenger constraints so that the overall solution creates less impact overall.

11:15 – 12:00 – Session 2-4
Vendor Presentations

13:30 – 14:15 – Session 2-5
Airline Operations Feedback Control with KPIs
Robin Reidel, MIT, USA

In this joint study by MIT, Carmen Systems and Lufthansa, we model the operations of an airline as a feedback control system. We develop a set of parameters or Key Performance Νndicators (KPIs) that characterize the inputs to the system (e.g. flight schedule) in terms of their correlation with the outputs of the system (e.g. delays). These KPIs will allow planners and operations controllers to quantify the operational consequences of their decisions. The approach also enables the use of powerful techniques that are commonly used in control theory. The KPIs are being developed with historical and simulation data.

14:15 – 15:00 – Session 2-6
Intertemporal Valuations, Product Design and Revenue Management
Guillermo Gallego, Columbia University, USA

We study inter‑temporal choice models where customer valuations, of a product or service delivered at the end of the booking horizon, evolve over time. We show that for a wide range of capacity levels, selling partially refundable fares (or equivalently, selling call options on capacity) results in significantly higher expected revenues than using low‑to‑high pricing (e.g., with booking and overbooking limits).  The use of options allows customers to self‑select products (including options) without the need of imposing fences like Saturday night stays. We analyze the fluid model in detail and show that the solution of the fluid problem is asymptotically optimal for the stochastic problem. We extend our analysis to multiple periods and determine the optimal time to issue and to exercise options.

15:30 – 16:15 – Session 2-7
Analysis of the Short-haul First Wave at British Airways
Alex Ross, British Airways, UK

Like most large network airlines, overall short-haul/domestic operational performance at British Airways is very dependent on effective planning and delivery of the 1^st-wave at base and outstations.   We present a simple, high-level means of illustrating the key levers impacting 1^st-wave performance and discuss the initiatives underway and challenges faced in raising this performance.

16:15 – 17:00 – Session 2-8
Evaluating Robustness with Support Vector Machines
Michael Mederer, Lufthansa, Germany

The evaluation of Schedule Robustness and corresponding KPIs are typically done by stochastic simulation or analytic distribution models. The complexity of quantitative validation of the modelling results, due to the purely deterministic reality makes it necessary to replace the empirical data by the simulation outcome. The output of these models is then used for comparison of robustness measures. This presentation introduces an approach of using Support Vector Machines and Regression Trees to overcome the disadvantage of generating a synthetic environment. It is an introductory work that reduces the problem of a robustness forecast to the statistical knowledge of certain KPIs.

Thursday – September 29

08:30 – 09:15 – Session 4-1
The Price is Right
John Lancaster, Manugistics, USA

Price optimization and price-sensitive revenue management in low-fare markets presents the P/RM practitioner with new forecasting problems. These include but are not limited to endogeniety, competitive actions, sparse data and cannibalization. This has led many airline OR experts to be suspicious of market response modeling. While it is true that the airline industry does present significant issues, similar problems have faced and solved in the retail and CPG world. This talk focuses on how these techniques were borrowed, adapted and successfully implemented in two separate price-sensitive revenue management systems and the implications for the next generation of airline forecasting.

09:15 – 10:00 – Session 4-2
New Generation in Demand Forecasting
Shankar Mishra, Ben Vinod, Richard Ratliff, Sabre, USA

Demand forecasting is an integral component of the revenue management process. Improving the accuracy of demand forecasts translates into more accurate inventory controls and hence higher incremental revenues. Airlines that control inventory by origin and destination frequently embark on forecasting demand directly, typically at Origin, Destination, Itinerary & Fare-class level, from passenger name record data. This poses some unique challenges from data perspective - volume and sparsity, as well as the process perspective – natural bias between shopping & forecasting processes. With the growth of online shopping, global distribution systems like Sabre have access to customer shopping data that can be leveraged effectively for forecasting demand at an O&D level. By aligning the demand forecasting process with customer shopping process, this next generation in revenue management demand forecasting methodology based on customer preferences, offers significant improvements in forecasts over current techniques. This presentation will provide an outline of the proposed methodology and related numerical results.

10:30 – 11:15 – Session 4-3
Innovation in Recovery Optimisation
Alan Dormer, CTI – Melbourne, Australia

CTI has been working with Qantas for over 10 years building and developing operations control and crewing solutions. A recent addition to this portfolio is recovery optimisation. There are a number of issues that arise from solving this problem. These include: 1 Solving the operations and crewing problem simultaneously. 2 How to improve decision making rather than taking the decision away from the user.  3 Dealing with external issues that the system may not (or cannot) be aware of. The paper will discuss these issues and how they relate to a practical solution of the problem.

11:15 – 12:00 – Session 4-4
Aircraft Maintenance Planning and Control Simulation
Massoud Bazargan, Embry-Riddle Aeronautical University - Daytona Beach, FL

This study presents a simulation model for the aircraft maintenance operation at AirTran’s maintenance station in Hartsfield-Jackson International Airport in Atlanta. The model involves various characteristics and behaviors of aircraft maintenance system, including gate assignment and its effects on the required number of mechanics, un-scheduled aircraft maintenance, scheduled Remain Over-night (RON) maintenance, and the constraints affecting re-source availability, ground time, and spare parts. Optimization modules of the simulation software generate an optimum shift schedule of mechanics thus increasing the existing system efficiency and matching mechanics to increased workload in the future.

13:15 – 14:15 – Session 4-5
Integrating Revenue Management and Fleet Assignment
Barry Smith, Sabre, USA

Fleet Assignment Models are used by many airlines to improve schedule profitability by matching capacity and demand. Schedule profitability is also affected by the subsequent processes of pricing and revenue management. The performance of fleet assignment models (FAM) can be improved by anticipating these effects. We’ll review approaches to incorporating revenue management into FAM. In particular we’ll review recent modeling improvements to incorporate O&D revenue management and restriction-free pricing into FAM.

14:15 – 15:00 – Session 4-6
An Internet-based Pricing and Booking Tool
David Post, SigmaZen, Germany

This presentation introduces a pricing and booking tool that airlines are using for the disposal of distressed inventory. The tool is targeted at time-flexible airline travellers and outsources the product specification to the customer. The tool generates attractive incremental revenues but an even greater benefit may be the creation of a rich data set of revealed customer preference data. This is generated from the high numbers of customer interactions as well as the results of the online questionnaire offered to purchasers. Furthermore, this data set may be used to develop sophisticated pricing algorithms.

15:30 – 16:15 – Session 4-7
Inventory Control ‑ Advances to Maximize Network Revenues
Umit Cholak, Sabre, USA

This presentation is about Sabre's new generation in reservations inventory control for maximizing revenues.  Flexibility in implementing different types of inventory controls based on market strategy and tactics is a requirement for the success of any revenue management program.   This new generation of inventory control that provides greater flexibility than traditional legacy systems was made possible by migrating inventory from TPF to an open systems environment.  A range of inventory control alternatives and availability calculation methods based on the desired level of sophistication of an airline to control seat inventory and performance of the new system are reviewed.

16:15 – 17:00 – Session 4-8
Increasing Airline Margin by Optimizing Corporate Contracts
Julian Pachon, Navitaire, USA

While corporate travel represents less than twenty percent of the major airlines’ travel load it accounts for more than eighty percent of their operational earnings. To remain competitive in the industry, airlines must manage their contracts with corporate customers in the most efficient manner while at the same time maintaining a high level of corporate customer satisfaction. In this presentation I will describe the corporate contract optimization problem, propose a solution approach and show numerical results based on real-world case study for major international carrier.

Friday – September 30

08:30 – 09:15 – Session 5-1
Does Reducing Crew Premium Really Reduce Costs
Alex Burger, Qantas Airlines

Reducing the non-productive expense (premium) in crew rosters is a major driver for investment in pairing systems. But how can we be sure that a reduction in actual crew costs was a result of better planning, rather than some operational anomaly. More precisely, how do we convince our Finance managers and other sceptics that planned premium really is a valid KPI that can yield cost saving benefits. This paper examines how these questions were addressed during a recent crew systems benefits analysis at Qantas.

09:15 – 10:00 – Session 5-2
Using Revenue Management Methods in Fleet Assignmt
Jutta Rockmann, Lufthansa, Germany

Fleet assignment is done by many airlines in order to maximize the overall profitability of the schedule. In the tactical timeframe from several months until very close to departure, the optimization process is based on a fully dated schedule and on demand data from revenue management. Key success factor in this process is the accuracy of the revenue forecast for each flight and for every aircraft type. While conventional methods used for calculating revenue take a single yield per compartment, Lufthansa developed a method using a standard EMSR-approch from revenue management and demand data on booking class level. First results show that the quality of fleet assignment increases significantly.

10:30 – 11:15 – Session 5-3
Operational Impact of Predeparture Sequencing
Wolfgang Hatzac, ATRiCS, Germany

More and more airports plan to approach the capacity problem by means of Predeparture Sequencing, also known as Departure Management (DMAN). In that process, optimised target offblock times (TOBTs) are assigned to flights in order to achieve an optimised departure sequence at the runways. But how do factors such as taxi time variation, runway topology and resource conflicts at the gates affect the expected benefits in throughput, taxi time and takeoff punctuality? The answer is presented on the example of a major European airport, identifying critical success factors and introducing possibilities to alleviate their impact.

11:15 – 12:00 – Session 5-4
Constraint Programming in Airline Optimization
Stefan Karisch, Carmen Systems

Constraint Programming (CP) emerged in the 1970s within the field of artificial intelligence, and research over the last 10 years has looked at the integration of CP with traditional Mathematical Programming (MP), and its application to optimization. In this presentation, we will review our own experience with combining CP and MP approaches for solving problems arising in airline scheduling and operations.

13:30 – 14:15 – Session 5-5
Applications of new class of genetic algorithms
Andrzej Slodownik – LOT/GE, Poland

As a result of developing new methods for solving multidimensional optimization problems of NP-type, a new class of very effective genetic algorithms has been developed. Their applicability for schedule optimization problem has been tested .The results of tests of effectiveness of the new algorithms for fleet scheduling problem and possibility of extention of their applicabilty for other areas is presented.

14:15 – 15:00 – Session 5-6
A New Approach for Solving the Probabilistic Nonlinear Seat Inventory Control Problem
Earl Barnes, Georgia Tech, USA

The seat inventory control problem is to maximize expected revenue from future scheduled flights by allocating seats to different fare classes. We solve the probabilistic version using the Langrangian dual formulation. We show that the Lagrangian dual can be solved by using a trust region based on Newton’s method, where the gradient and Hessian are computed efficiently by exploiting information from the primal representation. Special cases and schemes to adjust the dual and primal variables are discussed.

Call for papers

The 45th Annual Symposium of the Airline Group of the International Federation of Operational Research Societies (AGIFORS) features a comprehensive program covering the whole spectrum of airline operations research. This year’s symposium will be held from 25 – 30 September 2005 in Guaruja, Brazil.  Contributions to the technical program are solicited in the following areas, but are not limited to, flight scheduling, pricing, yield management, maintenance and engineering, crew, ground resources, and finance.  Papers covered by active Study Groups, as well as other topics not covered by active Study Groups, are sought.  Airline, consultant, academic, and industrial research representatives are all encouraged to attend.

Presentations are 30-35 minutes, with 10-15 minutes for Q&A.

Symposium proceedings will be published by January 2006.

Abstract Submissions - August 1, 2005

Please send all submissions via the online form.

Submit abstracts online (maximum 100 words). Include a title and identify the corresponding author.  For each co-author, include full name, affiliation, complete address, telephone number, FAX number, and electronic mail address or URL.

Paper/Presentation Submissions -  September 1, 2005 

Corresponding author provides both an electronic file submission of the presentation/paper for inclusion on the CD-version of the Proceedings and a photo-ready copy for the printed Proceedings.

Please note the following conditions for inclusion of your paper in AGIFORS 2005:

• At least one author of an accepted paper must attend the Symposium to make the oral presentation.
• Each attendee must register and pay the registration fee.
• Each attendee is responsible for making all hotel arrangements.

Tassio Carvalho
AGIFORS Symposium Technical Program Chair
Phone: +1 817.931.9273
E-mail: technicalchair@agifors.org