Log in

SSP 2021 Technical Presentation Abstracts


Integrated instead of Sequential Optimization - A Novel Approach to Solving the Multi-Hub Network Planning and Scheduling Problem

Marius Radde, Ralf Schneider, Peter Lietz, Judith Semar

Lufthansa Systems


The airline network and schedule design process is characterized by sequential problem-solving using highly specified optimization tools and a large amount of tedious, semi-manual work steps along the way. An integrated optimization approach combining destination selection, frequency planning, time optimization and fleet assignment to generate profitable (multi-)hub networks taking into consideration O&D demand and competition has always been infeasible due to the large number of possible solutions. Long run times prevented the application of such optimization approaches in real business context.

Together with one of the world’s largest airline groups we developed an innovative optimization approach that solves the network planning and scheduling problem in an integrated manner for the first time. Our new optimizer harmonizes supply and demand while taking operational and commercial restrictions into account. This results in an operationally feasible schedule that is optimized to meet demand and generate maximum network profitability. Thanks to a highly efficient computational implementation we evaluate hundreds of millions of schedules within a very large-scale neighborhood search.

In this presentation we will give you insights on the implementation of this novel optimization approach. We will show how given solutions are evaluated regarding their quality, how new solutions can be automatically created from-scratch, how we steer the optimization in the right direction and – of course – how the approach performs on real-world applications.


A Completely New Platform to Harmonize Lufthansa Group’s Network Planning

Fabio Ghielmetti, Susanne Schwanz

Lufthansa Group


Lufthansa Groups has three Hub airlines, Lufthansa itself, Austrian Airlines and Swiss. All three airlines developed over time sophisticated network planning and scheduling processes. To increase speed to market it was needed to ease and harmonize communication and processes within the Lufthansa Group while at the same time not neglecting special requirements of each airline. 

In an agile project set-up started with a Design Thinking workshop a Minimum Viable Product has been created and went live after just 7 month development time. The new web-based tool allows the planner to create, edit and retrieve schedule scenarios and to directly create the business case out of it. This is now possible not only by the a new user interface, but also by choosing a new way of modeling market share effect of schedule scenarios using a delta approach.


An Airline Scheduling Perspective on Urban Air Mobility

Sergej Bukovac

Lufthansa Systems


News of Urban Air Mobility start-ups is generating excitement around the world. However, Urban Air Mobility (UAM) still needs to find its space, purpose and identity in the transportation industry context. It flies, but it is not an airline. It moves people but it is not a mass public transport. It overlaps with road, rail and air transport and provides services in multiple catchment areas (local, metropolitan and regional). But does it challenge or complement any of the existing transportation modes? In this presentation we explore the significance of UAM emergence on airlines and airline planning and scheduling. Multiple factors will decide if airlines need to adapt, adopt or compete with the UAM in the future. As always, planning and scheduling areas will be tasked with the execution of airline response. What can airlines do right now to prepare for this? Also, what are the potential learnings from airline planning and scheduling that can be transferred to the Urban Air Mobility operators?

Improvements to Market Size Estimation Using Supplementary Data

Anupama Lakshmanan, Ramakrishna Thiruveedhi, Naveen Kumar CV

Sabre Labs


The estimation of the future demand for airline markets, where the market demand is defined as the number of passengers traveling from an origin airport to a destination airport, is an important problem in the airline industry. The use of time series models to estimate the demand based on historical data has been well established. We also use time series models like XGBOOST to estimate the market size for all existing markets. However, these models do not use supplementary information which can help improve the estimates significantly. In this work, we use shopping data along with economic data like GDP (Gross Domestic Product) to improve the forecasts of market size. The results show an improvement of more than 5% Absolute Percentage Error for most markets.

Dynamic Scheduling

Shahram Shahinpour, Ramakrishna Thiruveedhi

Sabre Labs


Schedule is the most important product of an airline. Commercial airlines carefully plan their schedule to ensure it is operational and profitable while adhering to safety and regulatory rules. A well-planned schedule is feasible and reliable while making the best use of airline resources. Any disruption caused by weather conditions, mechanical issues, reduced availability of crew or ground resources like gates reduces actual profitability. Thus, airlines need reliability in their schedule and maximizing planned profitability is not sufficient anymore.

Additionally, the current global pandemic has led to increased demand uncertainty and airlines are operating in an uncharted territory. To navigate this complex business environment, airlines rely on dynamic scheduling and last-minute changes in capacity to manage sudden fluctuations in demand. How can airlines implement dynamic scheduling while making sure all operational constraints are met?

In this talk we focus on incorporating operational feasibility and reliability in schedule development process. First, we propose a solution to integrate gate planning into schedule design. We measure the benefits using a schedule planning case study. We then propose an approach to balance profitability and reliability and show various levers available to an airline to achieve this.

Flight Numbering Automation

Jeff Warren

Sabre Labs


We will describe the flight-numbering policy at a major airline and how Sabre automated much of the work required to implement that policy. We will address the airline's need for long-term numbering consistency, day-to-day numbering consistency for new flights, respect for flight-number ranges by operating carrier and service type, the resolution of callsign conflicts, and the airline's need to conserve flight numbers through limited re-use of flight numbers within the same day. We will describe the challenges we have faced in automating flight numbering, including enforcing business rules, balancing competing requirements, and the immense computational costs of optimizing integer programming models.

Choice-Based Airline Schedule Design and Fleet Assignment

Vikrant Vaze - Thayer School of Engineering 

Chiwei Yan - University of Washington

Cynthia Barnhart - MIT


We study an integrated airline schedule design and fleet assignment model for constructing schedules by simultaneously selecting from a pool of optional flights and assigning fleet types to these scheduled flights. As passenger demand is often substitutable among available fare products between the same origin-destination pair, we study an optimization model that integrates within it a passenger choice model for fare product selections and solve it using a novel decomposition approach. We conduct detailed computational experiments using two realistically sized airline instances to demonstrate the effectiveness of our approach. Under a simulated passenger booking environment with both perfect and imperfect forecasts, we show that the fleeting and scheduling decisions informed by our approach deliver significant and robust profit improvement over benchmark implementations and previous models in the literature.

Use of Shopping and Booking Data to Calibrate QSI Models

Jim Barlow, Adam Seredynski

Amadeus


Quality of Service index (QSI) models are typically used by schedule planning solutions to measure schedule attractiveness. These models are typically calibrated using aggregated data such as MIDT and traffic statistics. However, use of such aggregated data often permits the true attractiveness of schedule effects to be disguised. For example, when measuring the attractiveness of different aircraft types from aggregated data, preferences for larger aircraft types are often confused with the benefits of larger capacities and customer preferences for lower prices. Understanding the details when customer make their individual itinerary selections provides insight into the customer choice process. This paper demonstrates the value of supplementing aggregated data with disaggregated data from individual shopping sessions to derive better estimates of QSI parameters. Applications of the results from this approach are also discussed.

Mining the Airline Schedule Data: Strategic View on Fleet Deployment and Mission Diversification

Ahmed Abdelghany, Vitaly Guzhva

Embry-Riddle Aeronautical University


In this research, we study the diversification of the fleet deployment in the airline schedule. The diversification is measured using the Differentiated Product Concentration Index (DPCI). A higher value of the index indicates that the airline’s scheduled capacity is concentrated over few fleet types, and vice versa. The measure is calculated for each airline worldwide for the period of 2000-2021. Thus, it can be used to understand and compare airlines’ strategic fleet deployment decisions under major industry events (e.g., higher fuel prices, airline merging, pandemic, etc.). Finally, the proposed index is compared against the mission of the airline, which is measured in terms of the diversification of the routes in the schedule.


Frequency Regulation: Approaches and Potentials

Felix Presto, Volker Gollnick - Hamburg University of Technology

Klaus Lütjens - German Aerospace Center


Projected emissions and delays caused by increasing air traffic require the assessment of new capacity managements measures. Current slot allocation mechanisms, that rely on grandfather rights, decrease system efficiency by amplifying frequency competition among airlines. A potential solution is the regulation of frequencies (i.e. the number of daily flights) between airport pairs. That way, airport and airspace capacity is regulated in an integrated way to incentivize airlines to deploy larger aircraft. Taking the EUROCONTROL-area as a use case, we develop and analyze different frequency regulation approaches. Efficiency gains, in terms of the potential to reduce emissions, operating cost and delays, are only possible if the economic-ecologic performance of available aircraft types and seasonal demand fluctuations are considered. Regulatory implications and challenges are discussed.


Fleet Assignment and Bank Structure Integration in Airline Scheduling Problem

Enis Ciftci, Vildan Özkır

Yildiz Technical University


In hub and spoke networks, flight arrivals and departures generally have a bank structure to increase connections among spoke cities. Main objective of this study is to determine that routes in the banks are planned with the correct aircraft type, right departure and arrival time so as to maximize the passenger flow of all destinations across the network and to fit the slot capacity of hub airport. In the first part, Airline bank optimization problem is introduced. A mathematical model is formulated for improving connection times among connecting flights by changing departure or arrival times of flights in the bank. The mathematical model aims to minimise the total waiting times for transfer passengers and generates schedules regarding slot constraints in the hub airports. In the second part, airline bank optimization problem is integrated with fleet assignment problem. Lastly, real-world case study is presented using the dataset from a major Turkish carrier.

An Optimization Model for Generating Aircraft Rotations Considering Various Operational Restrictions

Francisco López Ramos, Mohammad Kouhi

Vueling


The Covid pandemic along with the frequently changes in the restrictions has forced the Vueling airline to constantly re-schedule the flight network and generate operable aircraft rotations considering various operational limitations. To respond to this, an automatic and flexible aircraft routing process needs to be created giving to the network scheduling department the agility to test different scenarios and to make best decisions. To accomplish this end, a decision-making module, called Rotation Builder (RB), has been developed for planning the daily aircrafts rotations giving the a flight schedule and ops planning limitations while the possibility of small retimings is included. This module contains several chained optimization models that are formulated as Mixed-Integer Linear Problems (MILPs). Each MILP have a different objective function but holds a common set of operational constrains such as minimum turn-around-time, maintenance stop and buffer.


Case Study - Taking Network Forecasting Accuracy to the Next Level During the COVID Crisis

Reza Baharnemati - Amadeus

Kaori Bray, Lonny Hurwitz - Southwest Airlines


COVID-19 restrictions have significantly impacted air traffic and have dramatically reduced the number of people flying. Airlines are looking for ways to minimize their risk and maximize their revenue by investing in markets with the highest rate of return. To achieve this, airlines need to know where people are most willing to fly and what type of services they are willing to buy. To answer these questions accurately, having high-quality data plays a crucial role in the decision process. However, due to the dynamic nature of markets caused by COVID-19, traditional approaches to scheduling and planning based on historical data were not effective enough to generate accurate network forecasts. To tackle these challenges and maintain a high level of network forecasting accuracy, Amadeus and Southwest Airlines teamed up and utilized different machine learning approaches along with new data sources.

Powered by Wild Apricot Membership Software