Presentation Abstracts (subject to final approval)
Anna Valicek Papers
Modeling Crew Itineraries and Delays in the National Air Transportation System
Keji Wei, Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire, USA
Dynamic Programming Decomposition for Choice-Based Revenue Management with Flexible Products
Sebastian Koch, University of Augsburg, Germany
Chairman and CEO of British Airways
Alex began his professional career at American Airlines [Decision Technologies] in 1995, spending half his 10 years at the group with its travel technology arm, Sabre, in London, working with a range of European airlines. In 2000, he became a partner at Arthur D Little before setting up his own aviation consulting firm in 2002. In 2005, he joined Accenture as its head of aviation. In 2006 he founded Clickair, a Barcelona-based airline, merging the airline with Vueling in 2009 and becoming Vueling’s Chairman and CEO. Vueling was acquired by International Airline Group (IAG) in 2013, with Alex joining the IAG Management Committee. In April 2016, Alex was appointed Chairman and CEO of British Airways. He is also a visiting lecturer at the IESE Business School in Madrid and the ESADE Business & Law School in Barcelona.
Alex is 50 years old and originally from Bilbao in Spain. He has a degree in industrial engineering from Central Michigan University, an MSc from the Ohio State University, and Business Management & Administration degree from the Cox School of Business in Dallas. He is married with four children and lives in London.
Amazon Air Network Design and Fleeting
[Ken Wang Tribute Paper]
Amazon’s Middle Mile Planning and Research Science (PROS) team is responsible for the design and implementation of optimization models and algorithms to design the air network and better utilize the Amazon air transportation assets. In 2017, PROS developed a network design and fleet assignment model to plan the air network for fall 2017 and beyond. The model uses a mixed integer programming approach that simultaneously builds the flight schedule, assigns fleets to flights, and determines how best to flow all air packages throughout the network. This model provides both tactical and strategic solutions taking into account available aircraft, station and hub capacity, the use of point to point and hub and spoke configurations, and other operational restrictions. Amazon uses the model to select new stations, determine operating strategies at the hub, and to evaluate network topology.
Balancing Revenue and Load-Linear Objectives in Capacity-Based Revenue Management
RWTH Aachen University
Beyond maximizing revenue, secondary load-linear objectives, such as capacity utilization, play a significant role in the long-term success of a firm. In practice, automated systems maximize airline revenue by optimizing inventory controls. But the revenue-maximal solution may not align with secondary objectives. As a workaround, in practice, analysts manually overrule optimization. Such manual interventions are unlikely to achieve an optimal balance of objectives. Thus, we suggest including a load-linear secondary objective in the optimization model to calculate a Pareto-curve as decision support to the analysts. We formulate a bi-objective optimization problem based on the classic independent demand revenue management problem and present an enumeration approach to computing a complete solution set of optimal trade-offs. Furthermore, we provide a heuristic to approximate the optimal solution set and show its quality in a numerical study.
Conceptual Models of Demand for eVTOL and Thin-Haul Aircraft
Georgia Institute of Technology
Spurred by new battery technologies, many companies are developing prototypes for distributed electric propulsion aircraft with vertical take-off-and-landing (VTOL) capabilities. These piloted or autonomous aircraft have the potential to dramatically reduce commuting times in urban areas with congested roadway networks. Given the novelty of these aircraft, we conducted four focus groups to better understand potential travelers’ perceptions and willingness to travel in these aircraft and pay for flights. We examined two use cases: one for intra-urban operations using aircraft with VTOL capabilities and the second for thin-haul markets serving destinations up to about 300 miles with an aircraft having nine or fewer seats. Based on insights gained from these focus groups, we present two conceptual models of factors that will likely influence demand for these aircraft in intra-urban and thin-haul markets.
Determining Optimal Maximum Take-Off Masses (MTOMs) in an Uncertain Flight Network
[Ken Wang Tribute Paper]
RWTH Aachen University
In order to reduce costs associated with MTOM-based landing- or navigation-charges, airlines may certify their aircraft for a reduced MTOM. In a deterministic world without any flight plan deviations, cost-minimizing MTOMs can directly be determined as the solution to a so-called integer optimization problem. However, in the case of unpredicted aircraft shortages, the MTOMs of a deterministic model may yield poor solutions in practice, as the chosen MTOMs restrict the swapping of aircraft among the flights of the disturbed schedule. In particular, a highly reduced nominal MTOM may imply cost-intense recovery actions like additional refueling stops or even the cancellation of flights. Therefore, we apply methods from robust optimization to the problem of finding cost-minimizing MTOMs in an uncertain environment. Ultimately, we identify MTOMs that reduce unnecessary charges while ensuring flexibility in reassigning aircraft in the case of unpredicted deviations.
A Dynamic Scheduling Methodology
Airline schedules are typically the same from one day to the next; however, passenger demand fluctuates significantly from day to day. The airline’s revenue management system can capture some value on high demand days, but ultimately a lot of demand is unsatisfied. In this paper, we present a methodology for varying capacity according to demand, day-by-day and market-by-market.
Flight Efficiency Improvements Enabled By Cross-Border Arrival Management for London Heathrow
Harris Orthogon GmbH
The presentation outlines the concept and benefits of the Cross-Border Arrival Management (XMAN) procedure. XMAN and its supporting technologies have been successfully deployed by UK’s Air Navigation Service Provider (ANSP) NATS in 2014 for inbounds to Heathrow. It saves airlines app. 4.700 t in fuel and reduces CO2 emissions by 15.000 t per year.
Improving a Path Generation Model
To predict the market share of an airline or investigate travel opportunities for passengers, we need to construct all possible routes between origin and destination cities. We model the airline network with cities on the vertices and flights on the edges. The generated paths should reflect the paths that are actually booked by passengers. We use two measures for the quality of a path generation model: passenger coverage and redundancy. To improve this quality we minimize the number of redundant paths for a fixed level of passenger coverage using a genetic algorithm. This algorithm selects the connectivity properties: maximum connection time, maximum detour and minimum quality index. We would like to discuss the pros and cons of the quality measures and the selection of the connectivity properties.
Making Numbers Count
[Best Presentation Crew Management Study Group]
Crew resource management is one of the critical factors in ensuring the smooth operations of any airline. In most cases information flow during planning is sequential however it is imperative that effective feedback loops are incorporated. Effective analytical feedback enables planning team to adapt quickly to ever changing landscape and improve the planning processes and reduce cost. Data exploration is not simple task in an airline environment as data resides in multiple specialized systems with complex structures and dynamic relationship which makes it tricky when they need to be brought together. Join us while we take you through our data discovery journey and how it influences management of crew resources.
Measuring Value of Revenue Management
[Armando Silva Tribute Paper]
How much value are we generating in Revenue Management? What is the remaining revenue upside? How do we know we are making good Revenue Management decisions? What actions should we take to capture additional revenue opportunities? These are a few of the questions facing every manger in airline Revenue Management. The answers to these questions bring insights into areas that require managers' attention for continuous improvement. In this presentation we will introduce several value measurement methods and discuss their strengths and weaknesses and key data requirements.
Modelling Strategic Trade-Offs
University of Westminster
The Vista project (SESAR Exploratory Research programme) examines market and regulatory force impacts on performance in aviation, through the evaluation of KPIs. Vista models the current, 2035 and 2050 timeframes based on such factors and their potential evolution. These factors include technological, operational and procedural changes envisaged as part of next generation ATM systems, plus wider market forces and regulations. A key objective is to better understand the trade-offs between KPIs and to identify future synergistic and antagonistic effects: what can we expect to improve by 2050, and what to worsen? This presentation describes the challenges and benefits of building a holistic model, taking into account both the strategic and tactical phases of operation, and feedback between them. What is the best approach to gain insight into how airlines may be impacted and how they should develop strategies for the future that are responsive to change?
A Much Improved Fleet O&D Algorithm with Tight Revenue Cuts
[Ken Wang Tribute Paper]
We present a much improved fleet assignment methodology that incorporates origin and destination network effect. The methodology integrates a network revenue management (RM) optimization engine with a leg based fleet assignment model. RM optimization engine is used to produce linear approximations to revenue (cuts). To overcome the problem that those cuts are not tight, we introduced a non-linear transformation from fleet assignment variables to seat capacity. The resulted cuts are much tighter. This approach leads to significantly improved solution time and quality.
Multi-Dimensional Nesting – A conceptual introduction to a unifying RM model
[Armando Silva Tribute Paper]
Delta Air Lines
Optimizing the Baggage Levers to Maximize Ancillary Revenue
Ancillary revenue is redefining the aviation industry and is no longer perceived as an “LCC model”. Excess Baggage revenue is one of the largest ancillary contributors and the maximization of this revenue stream is based on optimizing three key pillars: policy, pricing and distribution. As part of this paper we walk you through our journey in building an opportunity model to identify the size of the prize and understanding the implications of pulling various levers across the three pillars. We analyze how passenger demographics and behavior influence policy; the importance of product and market alignment and how dynamic pricing can further boost revenues. Finally, we touch on the role of technology in the transformation of ancillary revenue and the successful bridging of the opportunity gap.
Passenger Online Reviews Processing Models and Practices
The advent of social media and online reviews are changing the way of hearing voice of passengers. It is the fastest way for gauging the pulse of passengers. What if the review is by a celebrity that can reach millions or that could lead to huge penalty on violations. Social media is a tool for attention seeking and processing them is key for successful passenger service. In this presentation we present our research work and discuss challenges on veracity, volume and variety of reviews. We will present practices used in processing passenger reviews like sentiment analysis, clustering models, Kano and Fodness & Murray models. We also discuss the underlying gaps on processing in an integrated way across travel eco system stakeholders. We will analyze few case studies and discuss how newer methods, architectures and practices like big data paradigms and natural language processing help disseminating insights at right time for better actions.
Probabilistic Decision Making and Operations: Why it pays to play games of chance
[Best Presentation Ops]
Decisions made in the absence of probability-based analytics will not just usually be wrong, they will consistently be wrong, especially when those decisions involve weather and ATC. One of the objectives of our Decision Support Vision for the Operations Control Center is to improve how we respond to weather events, specifically in terms of the determining the timing, level of risk, and specific actions to take. We need to move away from deterministic decision making that is primarily ad-hoc and experience-based to a more data-driven process that is built on predictive analytics and accounts for uncertainty. This talk looks at some examples where probabilistic decision making could be applied, some examples in the industry of where these techniques are already being applied, and finally, our current efforts to develop this capability internally.
Recent Advancements in Effective Disruption Management
Airlines are faced with several types of disruptions that affect regular operations such as curfews, weather, unplanned maintenance and so forth. We present two major improvements on how operational considerations are factored into the schedule recovery process. First, we look at a passenger flow model that incorporates passenger re-accommodation decisions during schedule recovery. The solutions generated as a result of this improvement reduces the impact to passenger flows in the airline network and reduce overall passenger inconvenience. Second, we look at a new approach to rescheduling aircraft maintenance events considering various operational constraints. Finally, we look at improvements for handling routing based minimum ground times during the schedule recovery process.
Resilient Airline Operations
The main challenge for airlines today is to be able to anticipate and prepare for operational disruptions and quickly and effectively recover when they take place. Disruptions result in deviations from the nominal airline operations plan and can cause, among others, flight delays, passenger discomfort and disturbances to fleet and crew schedules. The costs and negative impact these disruptions have for airlines and their passengers are very significant. In order to enable resilient airline operations, representatives from 15 major airlines were interviewed to identify the most important challenges and describe how they impact airline operations. Several areas of highest interest in developing Decision Support Solutions (DSS) emerged and are divided in two main categories: Disruption Readiness and Disruption Recovery. To tackle those challenges in a holistic way, we propose Resilient Operations model that consists of analytical and optimization modules: Disruption Readiness and Disruption Recovery modules, respectively. Disruption Readiness module focuses on delay propagation and root-cause analysis by using data analytics. Disruption Recovery module focus on a holistic optimization by using decomposition methods and distributed optimization techniques that in a coordinated way integrate domain specific optimization algorithms by taking into account all key parameters, objectives and constraints relevant for resilient airline operations. The proposed methodologies and initial findings will be described and discussed in this presentation.
Revenue Management for Airline Solution Providers
Envisage an airline that aims, in a highly competitive market space, to maximize its revenue by offering a finite number of products and having a fixed capacity of resources consumed by those products over a finite period of time. Similar analogy may be used for products and services offered to airlines by solution providers. While analogy is similar, approach to solve that problem is slightly different. The Airline Solution sector covers products such as charting, flight planning and dispatch, fleet management, maintenance management, and crew management, as well as business consulting, technical support, call center operations, and software development. Firms may choose to purchase, rather than perform, these business functions to reduce costs, to mitigate risk, or simply to focus on their processes that provide marketplace differentiation. Developed heuristic model improves revenue streams for Airline Solution Providers by determining a list of offered products to aviation customers across all market segments and regions. Within this context, product offerings consist of standalone products as well as packages including upsell and cross-sell opportunities of most adjacent products and ancillary services. Model includes, among others, product characteristics such as pricing, strength, adjacency, and availability on one side and market characteristics such as capability needs, budget availability, business process alignment, technology compatibility, and competitive threat on the other side. Proposed approach and obtained results will be presented.
Strategic Fleet Replacement and Fleet Assignment Model to Determine Acquisition Strategies
Fleet replacement decisions are long-term strategic decisions with large economic and efficiency impact. This study presents a fleet replacement and fleet assignment model in order to determine optimal acquisition strategies. The model economically evaluates different type of aircraft obtained by various acquisition and replacement strategies such as financial leasing, purchasing, wet leasing and dry leasing based on the acquisition and sustaining costs that incur during the operation of the aircraft. The model is extended in order to integrate replacement decisions with basic fleet assignment decisions in order to optimize replacement decisions considering improved demand, network and utilization parameters. Solution methods and computational results pointing out the critical parameters are discussed.
The Air Cargo Load Planning Problem
FZI Research Center for Information Technology
The production planning in an air cargo terminal is a major operational problem for cargo carriers. For each flight a challenging puzzle has to be solved to determine the ULDs to build, when and how to build them, as well as where to load them into the aircraft.
We evaluated our approaches on more than 500~flight instances. The numerical results show that the developed models are suitable to automatically generate cargo load plans for flights and can significantly improve the quality of the load plans.
The Bid Price - Swiss Army Knife of Revenue Management
[Best presentation Revenue Management Study Group]
The bid price or a similar concept is at the heart of all Revenue Management methods, although it is not always visible and sometimes called differently. Despite its widespread use and the large body of scientific research there are many misconceptions about the bid price, especially among practitioners. As a result it is not used to its full potential, but instead often misused to pursue goals it is neither meant nor suited to achieve. Besides a short reminder about its definition, this presentation focuses on the economic meaning and some interesting properties of the bid price. The second part covers implications for RM practice, including a number of potential applications apart from the core RM task of computing booking class availability.
The End of RMS As We Know It? (Deep) reinforcement learning for airline revenue management
[Armando Silva Tribute Paper]
Reinforcement learning (RL) is an area of machine learning concerned with how machines take actions in order to optimize a given reward (e.g. revenue) by interacting with its dynamic environment. Some well-known recent applications include self-driven cars and machines playing games better than humans (e.g. chess and go). One of the main advantage of this approach is that there is no need to explicitly model the nature of the interactions with the environment. In this work we present a new airline revenue management optimizer based on reinforcement learning. The model does not need neither a demand forecasting nor customer modeling (i.e. to estimate wiliness to pay) to work. It is theoretically proven that RL will converge to the optimal solution, however in practice, the system may require a lot of data (e.g. thousands of years of historical bookings) to learn the optimal policies. To overcome these issues, we present a novel model that integrates domain knowledge powered by a deep neural network trained in specialized hardware. The results show very encouraging results with different numerical /simulated scenarios. We believe this opens the door to a new generation of revenue management system that could automatically learn by interacting with the competitors and customers, so it can react much faster to changes in the market conditions.
The Future of Aircraft Boarding
DLR German Aerospace Center
The aircraft trajectory on the day of operations depends on efficient arrival/departure procedures and enroute performance, but also on a reliable and efficient turnaround. The aircraft boarding as a critical turnaround process is driven by the passenger’s experience and willingness or ability to follow the proposed procedures (e.g. late arrivals, no shows, amount of hand luggage, status passengers). The presentation will focus on a reliable passenger boarding model, results of a measurement and validation campaign, and investigations into infrastructural changes (Side-Slip Seat). Two new topics in the context of aircraft boarding will be additionally introduced: the online prediction of the boarding time using sensor information from the connected aircraft cabin and the SeatNow concept, which addresses operational improvements if the current standard call-in boarding procedure is replaced by a dynamic seat allocation process.
Towards the Optimal Balance Between Operational Planning and Recovery
Georgia Institute of Technology
It is widely accepted that proper planning is essential to good airline operations. However, we live in a stochastic world. Thus, we must either incorporate buffers in the planning phase to hedge against the subsequent effects of disruptions, or develop tools for re-planning or recovery. In this presentation, I will review the state-of-the art in both operational planning and operational recovery, and propose a methodology for determining the optimum balance between operational planning and recovery.
Zero based Network Planning Model
Many existing network planning tools start from the current schedule and build on it, meaning the business doesn’t know if solutions are a local or global optimum. There is also significant uncertainty in factors influencing the airline market and therefore a range of assumptions and scenarios need to be considered in long term plans. BA has built a “zero based” long term planning model that captures the complexities of the aviation market in a way that enables the business to answer a range of strategic network planning questions and to validate decisions under different scenarios. This presentation will describe how the model has been developed, the benefit it is providing and the type of questions it is being used to answer.