|58th Annual Symposium
|Activity, Presentations and Speakers
|Welcoming Reception [sponsored by Boeing and Jeppesen]
|Opening Remarks and Introductions - AGIFORS
|Towards a Block Chain-enabled Aviation Ecosystem, John-Paul Clarke, Georgia Institute of Technology
|Predicting demand for commuter air-taxi service, Pr Laurie Garrow, Georgia Institue of Technology
|Tea Break [sponsored by Amadeus]
|From Revenue Management to Dynamic Offer Management, Jonas Rauch, Lufthansa
|Hub Connectivity: Discussion Of Measures, Competitive Profiling And Efficiency Evaluation, Zeliha Akca, Turkish Airlines
|Leveraging cognitive biases and the decoy effect to model sales conversion, Thiery Delahaye, Amadeus
|Business Luncheon [sponsored by WheelTug]
|Tail Assignment from routing to ops control, Mattias Gronkvist, Jeppesen
|Keynote by Ken Ejiri, Director Revenue Management, All Nippon Airways
|Vendor presentations: WheelTug, Sabre, OAG
|Tea Break [sponsored by OAG]
|Anna Valicek Paper Competition - Richard Cleaz
|An optimization approach for airport slot allocation under IATA guidelines, Nuno Antunes Ribeiro, CITTA, University of Coimbra Polo II
|Total Revenue Optimization with the Ancillary Marginal Demand and Ancillary Marginal RevenuTransformatione Heuristics, Adam Bockelie, MIT ICAT
|End of Technical Session for Day 1
|Predictive maintenance using aircraft sensor data anomaly detection. Dominic Wigmore-Shepherd, British Airways
|Airline Hangars Balanced Manpower Utilization – An Optimization Approach, Massoud Bazargan, Embry-Riddle Aeronautical University
|Vendor presentations: Hitit CS, T2RL
|Tea Break [sponsored by Hitit CS]
|Callable Products with Early Exercise and Overbooking, Pr. Guillermo Gallego, Hong Kong University of Science and Technology & Pusan National University
|Using Deficit Functions for Aircraft Routing and Crew Scheduling, Helman Stern, Ben Gurion University of the Negev
|Variable Pricing, Sergey Shebalov, Sabre Airline Solutions
|Housekeeping and logistics
|End of Technical Session for Day 2
|Business Luncheon [sponsored by FlightAware]
|Networking Event [sponsored by OAG]
|Networking Dinner [sponsored by Sabre]
|Big Data / Machine Learning at American Airlines Tech. Ops. Mei Zhang, American Airlines
|Quantum Computing for Airline Problems, Youngburn Hur, Sabre
|Efficient Route Planning for Air Cargo, Felix Brandt, FZI Research Center for Information Technology
|Tea Break [sponsored by T2RL]
|The Implications of Dynamic Pricing for Airline Revenue Management, Michael Wittman, Amadeus
|Business Luncheon [sponsored by Merlot.aero]
|Data Analytics combined with RM strategies is shaping the future of RM at Air Transat, Aimé Kamgaing Kuiteing, Air Transat
|Vendor presentations: FlightAware, Boeing & Jeppesen, Merlot.aero, OctoberSky & Gurobi
|Crew and Fleet Integration: Crew Pairing, Retiming and Aircraft Routing Combined, Waldemar Kocjan, Jeppesen
|Tea Break [sponsored by October Sky & Gurobi]
|The impact of competition on booking behavior in airlines, Daniel Hopman, Vrije Univeristeit Amsterdam
|[Best Presentation Crew] Pairing and Roster Optimisation in the Cloud, Olga Perederieieva , Merlot Aero
|End of Technical Session for Day 3
|Gala Dinner [sponsored by Amadeus]
|Data Science Applications to Improve Technical Operations, Rod Tjoelker, Boeing
|Bringing together Network scheduling and OCC tfor a more efficient tail assignment process, Benoit Robillard, Air France
|Tea Break [sponsored by Sabre]
|[Best Presentation SSP] Models to Support Scheduling Decisions in a Complex Network: Fleet Assignment Model as an Example, Yuxi Xiao, American Airlines
|A Field Experiment on Airline Lead-in Fares, Alexandre Jacquillat, Carnegie Mellon University’s Heinz College
|Study Group updates
|Best Presentations Award Voting
|End of Technical Session for Day 3
|Closing Business Lunch [sponsored by Amadeus and Sabre]
Hub Connectivity: Discussion Of Measures, Competitive Profiling And Efficiency Evaluation
Zeliha Akca - Turkish Airlines
Hub connectivity is an important multi-dimensional performance indicator to evaluate network competitiveness and decide on network improvement strategies. This research discusses different hub connectivity measures, important network dimensions and the relation to the perspective of the person measuring connectivity. From an airline perspective, a new connectivity measure considering commercial objectives is presented to benchmark the major airline hubs revealing strengths and weaknesses based on selected performance dimensions. Hub-and-spoke network model aims to maximize the coverage of network by using available resources more efficiently. In this respect, to evaluate the efficiency of an airline’s hub, this research provides a methodology with respect to resources and the resulting connectivity. Results are presented for major competitors and improvement strategies are discussed.
Airline Hangars Balanced Manpower Utilization – An Optimization Approach
Massoud Bazargan - Embry-Riddle Aeronautical University
Airlines on average spend 10%-15% of their total operating cost on aircraft maintenance checks. Many airlines conduct their light checks in-house. These checks are inherently very labor intensive. This study offers a mathematical model to help airlines with scheduling their in-house light maintenance checks to achieve a high and even utilization of manpower among all hangars. The model attempts to minimize the utilization imbalances by assigning light maintenance checks, during their feasible time-windows, to available and certified hangars on daily, weekly and monthly. It incorporates daily flight schedule, maintenance requirements for each fleet type, certification and availability of manpower at each hangar. Furthermore, the model highlights the utilization imbalances and thus provides some guidelines in terms of increasing/decreasing manpower capacities at each hangar. The model is applied to 3 US and 2 European airlines with encouraging results.
Total Revenue Optimization with the Ancillary Marginal Demand and Ancillary Marginal Revenue Transformation Heuristics
Adam Bockelie, Peter Belobaba - Massachusetts Institute of Technology, International Center for Air Transportation
Changes in airline business models over the last ten to fifteen years have led to a rapid growth in ancillary services and ancillary revenues. However, the development of revenue management models has not kept pace in this domain; availability controls are still designed to maximize ticket revenues, with ancillary revenues as an afterthought. A more comprehensive approach would favor booking policies that maximize total revenues. We propose an Ancillary Choice Dynamic Program for total revenue optimization that explicitly incorporates the revenues and passenger choice impacts of ancillary services in addition to ticket revenues. We use an estimate of conditional passenger choice probabilities to compute choice and ancillary-adjusted marginal revenues for booking policies through an Ancillary Marginal Revenue transformation, and we develop an Ancillary Marginal Demand forecasting model to estimate demand volumes. We combine the AMD and AMR frameworks as total revenue optimization heuristics for existing RM optimizers, such as EMSR. We discuss implementation challenges and use the Passenger Origin-Destination Simulator (PODS) to illustrate the performance of our approach versus traditional RM models. The results suggest that AMD and AMR can increase revenue for airlines by up to 2%. Finally, we discuss the potential for our dynamic program to be used as the basis of an offer generation system, leveraging the power of New Distribution Capability.
Efficient Route Planning for Air Cargo
Felix Brandt - FZI Research Center for Information Technology
Recent advances in algorithm engineering allow to find journeys in large public transport networks in a matter of milliseconds. However, these algorithms apply primarily to passenger trips in railway networks and are not directly applicable for finding air cargo journeys. We introduce a set of new extensions to these algorithms that incorporate typical air cargo restrictions like cargo compatibility with aircraft/airports, freedoms of the air, individual minimum connection times, and ad-hoc road feeder services. We present the results of our first experiments that allow us to find valid cargo journeys in less than 1 ms in large flight networks. As this is work in progress, we seek for any kind of community feedback, especially additional requirements for usage in practice.
Towards a Block Chain-enabled Aviation Ecosystem
JP Clarke - Georgia Institute of Technology
Airline operations is replete with regulatory requirements with respect to the documentation of events, especially those related to maintenance, repair, and overhaul (MRO). Currently, however, much of that documentation is done manually. Thus, the information contained within the documents cannot be easily verified and validated, and the associated data is not readily available for analytics. In this presentation, we will explore the use of electric documentation and block chains as a basis for verification and validation of MRO events, and as the foundation of an analytics and optimization structure that could lead to greatly improved operations.
Leveraging cognitive biases and the decoy effect to model sales conversion
Thierry Delahaye, Rodrigo Acuna-Agost - Amadeus
Most passenger choice modelling in the travel industry assume that customer actions are rational. Empirical evidence and past research, however, has shown that these assumptions do not always hold in practice, and that cognitive biases have a significant impact on choices and particularly on conversion. A typical such bias is the decoy effect. It is observed that introducing an additional alternative (the decoy) in a “menu” can increase the probability of some existing alternatives and even increase conversion; a behaviour in contradiction with the rationality assumptions (i.e., new alternatives should always decrease the probability of existing alternatives). In this presentation we show alternative approaches to model the decoy effect, including context-adjusted Multinomial Logit (MNL) and non-linear machine learning models. The models are then tested on two airline problems: a) flight sales conversion, where we study how the composition of a “flight menu” influences conversion. And secondly, on b) fare family choice, where we study how the composition of a fare family menu impacts passenger sell-up. For each, we focus on detecting quantitatively the decoy effect, and on leveraging it to better describe and predict passenger choices and conversion.
Callable Products with Early Exercise and Overbooking
Guillermo Gallego, Haengju Lee - Hong Kong University of Science and Technology & Pusan National University
Capacity providers such as airlines have traditionally increased revenues by practicing market segmentation and revenue management. However, they have left money on the table by neglecting to broker capacity between consumers with different willingness to pay. With the introduction of callable products, some consumers who buy capacity at low-fares grant the option to the capacity provider to recall capacity at a pre-specified recall price. These options in turn allow the capacity provider to buy back capacity to satisfy demand from high fare consumers when capacity is exhausted. Although the idea of callable products was introduced before for the special case of two fares, we make the procedure more operational by allowing multiple fares and restricting the option exercise policy to allow displaced low-fare customer to make alternative plans. Our model allows the service provider to keep their booking limit and overbooking policies in place. Our numerical study illustrate how callable products are win-win-win, providing additional revenues to the capacity provider, better service to high-fare consumers, and higher expected surplus to low-fare consumers who grant recall options.
Predicting demand for commuter air-taxi service
Laurie Garrow, Patricia Mokhtarian - Georgia Institute of Technology
Improvements in battery technologies offer the potential for dramatically lower operating costs for new classes of electric propulsion aircraft. By reducing aircraft operating costs, electric propulsion could transform both air and surface transportation. There is now widespread belief that on demand mobility (ODM) missions could be served by smaller electric propulsion aircraft with vertical take-off-and-landing (VTOL) capabilities that operate from vertiports or similar infrastructure. It is envisioned that these electric-VTOL (eVTOL) aircraft could provide air-taxi service for trips of two to four passengers between 10 to 70 miles within congested urban areas. In this presentation, we discuss results from a market survey conducted of commuters in five cities in the U.S. and identify potential market segments and market sizes for eVTOL flights for commuting purposes.
Tail Assignment from routing to ops control
Mattias Gronkvist - Jeppesen
The tail assignment problem assigns aircraft to flights from the day of operations and a few days or weeks into the future. Ops control is the process of deciding which schedule to operate and how to use the aircraft when operational disruptions happen. In this presentation we will discuss how a combined ops control and tail assignment system gives benefits compared to using separate systems. We will also show how integration with crew and flight planning can give additional benefits, and consequently how our tail assignment optimizer can improve airline efficiency and robustness throughout the fleet planning process.
Quantum Computing for Airline Problems
Youngburn Hur - Sabre
Quantum Computing has been in the news lately. Companies such as Google, Microsoft and IBM are investing millions of dollars into quantum computers. This talk will introduce quantum computing concepts, discuss why we might care about it, explain how quantum computing can be used for some traditional optimization problems and what Sabre is doing in quantum computing (in collaboration with academia and a leading quantum computing company).
A Field Experiment on Airline Lead-in Fares
Alexandre Jacquillat, Maxime Cohen, Juan Serpa - Carnegie Mellon University’s Heinz College
An airfare ladder includes a lead-in (the cheapest fare) and several sell-ups (higher-priced fares). Commonly, airlines match their competitors’ lead-in fares, regardless of differences in itinerary characteristics (e.g., scheduled time, number of connections). We partner with a leading global airline to challenge this long-standing practice. We design a field experiment that differentiates the lead-in fare during an eight-week period, affecting thousands of itineraries and dozens of thousands of bookings. We propose an experimental design that estimates the treatment effect by exploiting temporal and cross-sectional variation across three types of control groups. Our results show that lead-in differentiation increases revenue and yield, relative to lead-in matching, by 0.35 – 0.75 and 0.5 – 1.52 standard deviations, respectively (without decreasing market share). Next, we identify heterogeneous treatment effects using the causal forests method. We show that lead-in differentiation is most beneficial when the destination has high traffic, the airline has low market share, the trip is long, the trip departs early morning or in the evening, revenue comes mainly from low inventory classes, and the airline has many competitors.
Data Analytics combined with RM strategies is shaping the future of RM at Air Transat
Aimé Kamgaing Kuiteing - Air Transat
In the recent years, despite the growth of global demand for air transport and tourism, Air Transat has been struggling from a flood of new airline seats and holiday package offers from competitors. This has generated a war of market share in an ultra-competitive industry which puts pressure on prices. The old fashioned revenue management based on averages, intuition and personal experience was not sufficient anymore to be both competitive and profitable. There was a strong need of revamping Air Transat RM strategies and processes in order to blend data-driven strategies with intuition and personal experiences. This led in March 2017 to the creation of the revenue management system and inventory management team which is responsible for developing and implementing strategies and techniques to optimize and grow airline and holiday package revenues. How do you optimize revenue using revenue management strategies when your data is unreliable and your business model is different to those of legacy carriers? How do you improve RM analysts decision-making process? In this presentation, we will introduce a customized heuristic for seat inventory optimization and some functionalities of the flight inventory management tool developed in-house that provide powerful insights to RM analysts. Then we will discuss the benefits brought by these developments.
Crew and Fleet Integration: Crew Pairing, Retiming and Aircraft Routing Combined
Waldemar Kocjan - Jeppesen
Airline crew planning is usually divided into several problems solved separately. In this presentation we consider the crew pairing problem, the task of finding anonymous crew trips covering all flights. To add flexibility to that planning step we incorporate features from the previous planning phases: retiming aimed at time table adjustments, and dynamic aircraft rotation aimed at refining aircraft routing/tail assignments. Both features offer a way of improving pairing solution quality, but were previously not possible to combine. In this talk we will show advantages of the combined problem and share our experience from using the system with that level of integration.
Pairing and Roster Optimisation in the Cloud
Olga Perederieieva , Oliver Weide - Merlot Aero
Airline pairing and roster problems are hard to solve to optimality. However, near optimal approaches based on column generation are available and often used in practice. In order to solve such problems to desired quality in a reasonable amount of time, powerful computational resources are usually required. Such hardware resources are expensive to buy and maintain. Instead we propose to use cloud computing. Such an approach is cost efficient for medium size problems and allows to avoid hardware maintenance and data backup costs. Efficient parallel algorithms can be implemented to use all cores of a virtual machine in the cloud. When the number of cores on one machine is not sufficient, it is possible to use message passing interface for communication between machines and, as a result, use computational resources of all available machines. We discuss aspects of implementing such parallel algorithms and using them in Microsoft Azure cloud computing platform.
From Revenue Management to Dynamic Offer Management
Jonas Rauch - Lufthansa
While Airlines today increasingly depend on modern distribution of offers via the internet their actual offer creation systems are still mostly based on pre-internet technology. To fully leverage the new opportunities of modern data analytics tools and internet-based distribution airlines need to digitalize their Revenue Management approaches initially developed decades ago and still based on RBDs and statically defined products and prices. By transforming the current RM methods into RBD-independent modules will allow for a truly holistic Dynamic Offer Management that considers product design, pricing and distribution alike. In this presentation we want to discuss the basic setup to transform current Revenue Management approaches of airlines into a Dynamic Offer Management. To do so we will first shed light on our definition of Dynamic Offer Management as this term is still vaguely described both in academia and industry practice. We will then describe what major changes of Revenue Management approaches and systems are required to pave the way towards RBD- and filing-independent offer creation. Splitting up demand forecasting and customer oriented willingness-to-pay estimation is the foundation to develop both modules in a most flexible and scalable way. Based on this setup the implementation of dynamic pricing is enabled and gradual extensions can be added in the future for optimized ancillary pricing and dynamic bundling.
An optimization approach for airport slot allocation under IATA guidelines
Nuno Antunes Ribeiroa, Alexandre Jacquillat, António Pais Antunes, Amedeo R. Odoni, João P. Pita - CITTA, Department of Civil Engineering, University of Coimbra Pólo II, 3030-788 Coimbra, Portugal
Air traffic demand has grown to exceed available capacity during extended parts of each day at many of the busiest airports worldwide. Absent opportunities for capacity expansion, this may require the use of demand management measures to restore the balance between scheduled traffic and available capacity. The main demand management mechanism in use today is the administrative schedule coordination process operated by the International Air Transport Association (IATA), which is in place at the great majority of busy airports outside the United States. This paper proposes a novel multi-objective Priority-based Slot Allocation Model (PSAM) that optimizes slot allocation, while complying with the complex set of priorities and requirements specified by the IATA guidelines, as well as with the declared capacity constraints at the airports. It presents an efficient computational approach that provides optimal slot allocation solutions at airports significantly larger than has been possible to date. The model is applied to two Portuguese airports, a small one (Madeira) and a mid-size one (Porto) using highly detailed data on airline slot requests and airport capacity constraints. Results suggest that PSAM can improve the efficiency of current practice by providing slot allocations that match better the slot requests of airlines. Equally important, PSAM can also quantify the sensitivity of slot allocation decisions to the various priorities and requirements specified in the IATA guidelines.
Bringing together Network scheduling and OCC tfor a more efficient tail assignment process
Benoit Robillard - Air France
The tail assignment problem is a well-known topic in the operations research field. In Air France, the sequence of flights for a given aircraft is built by OCC a few days before Day of Ops, taking into account the flght maintenance planning. In order to break the silos and to improve operations efficiency, AFKL's OR team has built a common tool for both OCC and Network Scheduling departments, enabling for a more consistent schedule from construction to day of operations. We solve the tail assignment solution by spreading available ground time for a more robust schedule, but also by optimizing on estimated fuel consumption for each flight, while respecting operational contraints. The result is a transversal tool, used by both OCC and scheduling departments, and built with the assistance of many departments of the company for an accurate modelling of costs and penalizations.
Sergey Shebalov, Xiaodong Luo - Sabre Airline Solutions
Pricing Decision Support has become one of the major focus areas for both academic research and practical applications in recent years. We will discuss several key components of that concept from strategic fare optimization to dynamic pricing. In particular we will introduce the concept of Variable Pricing that allows simultaneously optimize both fares and allocations. We will present simulation results indicating significant revenue generating opportunities and describe several avenues for further development. We will also share our experience of implementing this concept and offer a few recommendations on the key factors that should be taken into account for successful adoption.
Using Deficit Functions for Aircraft Routing and Crew Scheduling
Helman Stern, Ilya B. Gertsbakh - Ben Gurion University of the Negev, Israel
We use deficit functions, DFs, to decompose a flight schedule into a minimal number of periodic & balanced aircraft chains. DF theory, developed in the 1960-70s has attracted more attention in bus than aircraft scheduling. Here we discuss its revival & show its use in integrating the aircraft & crew scheduling problems. A DF for each terminal has unit steps at flight departure & arrival times. The crucial use of DFs is to find the minimum fleet size & its chain decomposition, CD. To reduce the number of unbalanced chains, a Maximal Balanced Chain Decomposition problem is proposed. Edge disjoint Euler graph cycle covers are used to convert unbalanced to multi period balanced chains. The crew pairing problem is decomposed into separate sub problems, one for each aircraft, avoiding crew paxing and buffering schedule disruptions. Our approach is unorthodox compared to IMPs, but exposes the underlying solution space for aircraft CDs which may be illuminating.
Data Science Applications to Improve Technical Operations
Rod Tjoelker - Boeing
As the amount of data generated by aircraft, maintenance operations, and engineering continues to grow; advanced analytics opens the opportunity to leverage that data to improve operations. Machine learning and statistical modeling techniques applied to flight data, maintenance actions, and technical data enable analysts and engineers to more effectively use data and improve response time. These models enable monitoring data streams for events of interest and detection of anomalies. Analysis of maintenance data improves diagnosis of issues and when combined with analysis of sensor data, can create models to predict upcoming issues. Our work in applying data science techniques in the aviation domain requires working around several challenges due to quality and of the data and the complexity issues to be modeled, but is enabling solutions for improvements to technical operations.
Predictive maintenance using aircraft sensor data anomaly detection.
Dominic Wigmore-Shepherd - British Airways
Flight delays and cancellations due to unplanned maintenance issues cost airlines millions of pounds each year due to unplanned maintenance costs, compensation costs and various other expenses. Therefore, there is significant opportunity to improve both operational and cost performance by better identifying the issues causing unplanned maintenance events. Modern aircraft are equipped with thousands of sensors that monitor different parts of the aircraft. Utilising data collected from these sensors can identify issues before components fail, allowing engineering teams to maintain their fleet more proactively. In my presentation I will discuss an anomaly detection scheme I helped set up with British Airways Engineering department. I developed a generalised trending algorithm that can pick up anomalous sensor readings as soon as they start to deviate from normal behaviour. Any anomalous readings can then be flagged to an engineer to visually inspect the data via a bespoke dashboard. By varying the input parameters, the algorithm is customisable to different data streams, allowing for rapid deployment on a range of sensors. Since deployment, the algorithm has picked up multiple issues on aircraft sensors; these alerts have driven physical inspection of the parts, and in some cases prompted proactive maintenance of various aircraft components. Alerts tracking sensor readings for a vital A380 component have led to two instances of proactive maintenance before part failure could occur. By comparison with similar disruption events involving the same component, the estimated cost savings from this are roughly £300,000 per instance, due to avoidance of EU261 compensations costs for up to 450 passengers, as well as other associated disruption costs.
The Implications of Dynamic Pricing for Airline Revenue Management
Michael Wittman, Clement Trescases, Thomas Fiig - Amadeus
New distribution technologies will soon allow airlines to dynamically adjust ticket prices based on the characteristics of each shopping request. In this presentation, we discuss the theoretical and practical implications of such a next-generation dynamic pricing system. We first describe how dynamic pricing differs conceptually from traditional airline revenue management, and how airlines could see revenue benefits from providing targeted increments and discounts to filed fares. Results from the PODS revenue management simulator are used to demonstrate some key findings of dynamic pricing research. We then explore the calibration and validation of a dynamic pricing model in practice. Specific attention is paid to the segmentation of incoming requests and the development of the customer choice model used to propose dynamic price adjustments. We close by addressing several common practical concerns with dynamic pricing, including potential customer and regulatory reactions, the possibility of forecast spiral-down, and the risks of a race to the bottom.
Models to Support Scheduling Decisions in a Complex Network: Fleet Assignment Model as an Example
Yuxi Xiao, Ronald Chu, Luis Ochoa - American Airlines
After a merger with US Airways in 2013, the new American Airlines, operating in 9 hubs and flying to 350 destinations in 50 countries, became the largest airline in the world by its fleet size and passenger carried. In the presentation, we discuss a set of new challenges to effective scheduling brought about by the broad scale of the combined network. One major challenge is to solve a big network problem efficiently given a set of complex constraints. Another important one is to build practical models that streamline and automate manual business processes, while fully integrating into the schedule generation process. It is essential to understand the difference between optimality of the model solution and effectiveness of the results from the business standpoint (“acceptable optimality”). A certain level of gap to optimality is imperative for obtaining quick turnaround with a big problem.
Big Data / Machine Learning at American Airlines Tech. Ops.
Mei Zhang - American Airlines
Big Data and Machine Learning are two new additions to OR in last a few years. As more and more aircraft data become available, and machine learning algorithms and platforms mature, we have seen great potential in both of them in assisting in maintenance operations. I will show a few applications in predictive maintenance and performance analyses using big data and machine learning at American Airlines.