Copying of this publication or parts of it is only permitted under the provisions of the German Copyright Act of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. From my first steps in the research until the final technical preparation of the manuscript, he has always provided a steady and reliable support and encouragement, which was essential for me and my work.
Introduction
Overview
The timetable is presented to potential passengers and is the primary airline product that most influences the passenger's choice of airline (Gopalan & Talluri, 1998b). In addition, an airline's schedule affects almost all operational decisions, and on average, 75% of an airline's total costs are directly related to the schedule.
Objective
Structure
Airline Scheduling Process
Introduction
- Airline Scheduling
- Outline
Some of the subproblems are grouped together to build aggregate phases in the airline scheduling process. Furthermore, one of the most important tasks of DSS is to automatically solve subproblems of the airline scheduling problem.
Flight Schedule Generation .1 Problem.1Problem
- Network Design
- Frequency Assignment
- Flight Scheduling
- Solution Models
- Network Design
- Frequency Assignment
- Flight Scheduling
The external perspective in the studies reduces their applicability to the flight schedule generation problem. In the following, a selection of models is presented that can be used to support the flight schedule generation phase.
Aircraft Scheduling .1 Problem.1Problem
- Fleet Assignment
- Aircraft Routing
- Solution Models
- Fleet Assignment
- Aircraft Routing
Given this fleet schedule, the goal of aircraft routing is to find a feasible and profit-maximizing allocation of physical aircraft to flight paths (Barnhart et al., 1998). The complexity of the fundamental problem of daily fleet allocation is studied by Gu et al.
Crew Scheduling .1 Problem
- Crew Pairing
- Crew Assignment
- Solution Models
- Crew Pairing
- Crew Assignment
In the approach of Lagerholm et al. 2000) presented a feedback artificial neural network (ANN) to solve the crew pairing problem. In the approach of Gamache et al. 1999) a heuristic is used to find a good integer solution.
Integrated Models .1 Overview.1Overview
- Models
In general, the crew pooling problem is solved first to reduce the complexity of the crew deployment problem. In the approach of Barnhart et al. 1998), the fleet assignment and crew matching problems are still solved sequentially, but the fleet assignment model is improved to include the properties of the crew scheduling problem.
Summary, Conclusion, and Future Challenges .1 Summary.1Summary
- Conclusion
- Future Challenges
Airline planning is therefore one of the most important planning tasks of an airline. In reality, however, many of the inputs to the flight schedule problem are stochastic in nature (Day & Ryan, 1997).
Foundations of Metaheuristics
- Introduction
- Metaheuristic Optimization
- Design Elements of Metaheuristics
- Solution Representation and Variation Operators
- Fitness Function
- Initialization
- Search Strategy
- Selected Metaheuristic Optimization Techniques
- Local Search: Threshold Accepting
- Recombination-Based Search: Genetic Algorithms
- Summary
Thus, with each application of the operator, a neighboring solution in the search space is obtained. In many cases, the fitness function corresponds to the objective function of the problem behind it. This objective function expresses the quality of the current solution with respect to the goal we want to achieve.
In local search strategies, a new solution is iteratively chosen from the neighborhood of the current solution. The suitability of the solution is used to guide the search process to regions in the search area with high quality solutions (intensification). Some problem-specific adaptations of the metaheuristics are necessary to achieve an efficient search process.
Integrated Airline Scheduling
Introduction .1 Motivation
- Structure
- Data
- General Data
- Demand Data
- Supply Data
Better solutions can be achieved if boundaries between subproblems of the airline scheduling problem are further relaxed. This section focuses on the latter requirement and provides an overview of the data used in this study. Besides their location, the second most important characteristics of the airports for airline scheduling are their operating limitations.
This study assumes only one closure period per airport, regardless of the aircraft type. Because the focus of the integrated planning approaches of this study is on the day-to-day scheduling problem of airlines, it is assumed that (connecting) flights compete with each other when taking up space. B The appendix presents five different planning scenarios used for the experiments with airline planning approaches.
Schedule Evaluation .1 Overview.1Overview
- Market Size Estimation
- Forecasting Techniques
- Drivers of Air Travel Demand
- Gravity Models for Air Traffic Forecasting
- Gravity Model Development
- Itinerary Construction
- Overview
- Connection Building Rules and Parameters
- Parameter Calibration
- Itinerary Market Share Estimation
- Overview
- Setup
- Multinomial Logit Model
- Custom Model
- Evaluation
- Final Model
- Passenger Allocation
- Profit Estimation
- Summary
Market Size Estimation: Estimation of the total number of airline passenger demand between any two airports. Coefficients Pi j Ci j Bi j Di j. independent variables have no effect) can be rejected for each variable at the 5% level. 0.1] aPRSi Indicates the total route market share of the airline operating in the market.
From the total number of observations, a randomly selected set of routes serves as either a calibration data set (CS) or a validation data set (VS). Given the number of passengers on the route, calculating the overall profit of a flight schedule F is straightforward. The number of travel alternatives then determines the complexity of the schedule evaluation (Belobaba, 1987).
Sequential Approach .1 Overview.1Overview
- Solution Steps
- Fleet Assignment and Flight Scheduling
- Aircraft Maintenance Routing
- Schedule Optimization
- Solution Process
- Supportive Functions
- Schedule Initialization
- Integration
- Experiments
- Calibration
- Analysis
- Summary and Conclusion
- Summary
- Conclusion
The general constraints of the fleet allocation problem presented on page 18 are incorporated into this model. Consequently, the connectivity of the timetable is included in the opportunity costs. Thus, finding an Euler trip in G would result in a solution to the maintenance path problem.
The number of flights to be removed is determined as a percentage of popt∈[0,1]of the total number of flights in the schedule. The number of new flights is chosen based on a percentage of pnew∈[0.1]of the number of flights in the current schedule. The twnew=2·tw parameter controls the length of the time window for the fleet allocation step (see page 103).
Simultaneous Approach .1 Overview.1Overview
- Conceptual Design
- Overview
- Representation and Variation Operators
- Fitness Function
- Initialization
- Search Strategy
- Experiments
- Calibration
- Analysis
- Summary and Conclusion
- Summary
- Conclusion
In the first rotation, the last flight of the first day is from DUS to MAD. One advantage of the adaptive checking of the search operators is the reduction in the number of parameters that must be set for each metaheuristic. The more solutions, the more likely the search steps will find a better solution.
The following paragraphs analyze the obtained solutions and the GA solution procedure. The selection probability for recombination-based operators is the lowest at the beginning of the optimization. The continuous shift from recombination-based search to local search can be explained by different characteristics of search concepts.
Evaluation
- Comparison
- Experimental Verification
- Market Structure
- Number of Aircraft
- Number of Airports
- Fleet Types
- Summary
In terms of gain, not a single run of the sequential approach could achieve higher load factors than the worst run of the simultaneous approach. For scenarios A and E, the situation is reversed, which however results in the lower seat load factor of the sequential approach. The – compared to the simultaneous approach – higher standard deviation of the sequential approach is most likely the result of the combination of deterministic and stochastic search during optimization.
The number of skill evaluations required determines the effort of the techniques, making the concurrent approach a technique that requires less computational effort and time. As a comparison, they also include the results from the first attempts of the sequential approach and from the simultaneous approach, which. The comparison of the two approaches is made in terms of the profit from the business of the obtained solutions and the required computational effort.
Summary, Conclusion, Limitations, and Future Work .1 Summary.1Summary
- Conclusion
- Limitations
- Future Work
Analysis of the quality of the solution and the process of finding calibrated models gave satisfactory results. In the metaheuristic simultaneous approach, the operating profit corresponds to the fitness values of the processed solutions. Some of the shortcomings and possible future improvements to the timetable assessment process and airlines' approaches to timetable planning have been discussed in the relevant sections.
The conceptual design of the concurrent approach represents general design elements of a metaheuristic that is not limited to the search strategies used. The flexibility of the solution approach described in the previous section offers many starting points for further improvements. For example, factors such as the regularity of flights, the length of rotations per fleet or economies of scale at hub airports can easily be incorporated into the objective function.
Summary, Conclusions, and Future Work
Summary
As a result, preparing an airline schedule is one of the most important, yet most complex, planning tasks for any airline. A possible decomposition of the overall problem and aggregation of the sub-problem into planning stages is suggested on page 10. Since, in general, decomposing a problem reduces the interdependencies between decision variables, and a succession of solutions limits the flexibility of subsequent planning steps, only suboptimal or even unfeasible solutions to the problem can be achieved.
One of the two approaches to airline schedule optimization (presented in Section 4.3) follows the traditional scheduling paradigm of iterative and sequential solving of subproblems of the overall airline scheduling problem. The second airline scheduling method (presented in Section 4.4) is based on self-adaptive metaheuristic optimization, where complete airline schedules are processed at once. The feasibility of the simultaneous approach is further demonstrated by verifying its results for systematically changed planning scenarios (Section 4.5.2).
Conclusion
Future Work
To minimize the effect of these disruptions, stochastic elements can be included in the schedule estimation. Although the level of detail in which airline operations are represented is much higher in this study than in previous contributions, there is still much room for further improvement regarding practical considerations. As a consequence, the scheduling procedures presented in this study should be further extended to include slot constraints.
Until now, this study represents a theoretical framework; its applicability in the real world of airline schedules remains to be assessed. Therefore, the planning scenarios and all inputs used in this study should be replaced by existing data from an airline. This should also enable the use of planning scenarios from regions other than Europe, to which this study was limited due to the availability of data.
Appendix A
Aircraft Data
- Scenario A
- Scenario B
- Scenario C
- Scenario D
- Scenario E
Each scenario consists of airline-independent general data and the specific situation of the airline. For each scenario, these elements are chosen according to the tables in the next section, which have been selected to show various potential scheduling problems. In addition, because the goal is to compile an airline's daily schedule, competing flights from a randomly selected day are included in the schedule review process.1 The following table B.1 summarizes the five different scenarios, including the chosen day. for competing flights, the number of aircraft and fleets, and the number of airports available for the optimization process.
The time required depends on the number of flights and itineraries to be evaluated, which in turn depends on the number of competing flights. In order to perform a sufficient number of experiments, the number of competing flights given by the OAG schedules is randomly reduced to 10% of the original value. This reduction does not distort the basic results because this reduction is applied to all experiments and - to maintain a realistic estimate of passenger demand - the given market sizes are also reduced to 10% of their original value.
Appendix C
Calibration
- Sequential Approach
