Overview
The airline industry experiences growing congestion in the airspace over Europe, making problems more and more common and increasingly difficult to solve without causing new problems. Management of disruption occurring on the day of operation has therefore become an increasingly important issue for the airlines. In fact, all airlines suffer from operational disruption on a regular basis. Disruption shows itself in many ways, whether it be a small disruption, such as an inbound delay from Paris, a medium disruption, say an air traffic control strike in France, or a massive disruption scenario, such as severe weather conditions (e.g. fog or snow) at the UK’s largest airport, London Heathrow.
This project aimed at producing a decision support tool for the management of disruptions in an integrated fashion, taking the key resources into account (aircraft, crew; and most importantly perhaps, the passengers).
The project overall aim was to improve the efficiency of the European transportation system by assisting airlines to recover faster from disruptions. The outcome from using the tool would be faster recovery from disruptions, better service to the customers, fewer delays, fewer costs for the airline, and less unnecessary flying due to disruption.
More in detail, the objectives were:
- To develop a decision support tool for integrated crew and aircraft recovery on the day of operation.
- To jointly develop a prototype version of the integrated system, which would be accepted by experienced operations controllers to be used in managing the operation. It was planned that the prototype would be developed to a state where it was ready to be operational though may still require some productisation.
- Through methodological development enable the real-time solution of large-scale re-planning problems, whereby the methods and algorithms that are been developed would be useful in a number of other industrial cases, for example routing of vehicles, production planning in shipyards, and network planning in telecommunication.
At the beginning of the project the consortium decided against using a traditional 'waterfall' project methodology. Principally this was due to the ambitious nature of Descartes. A methodology was needed which would allow the requirements to be collated in an iterative manner, therefore minimising the risk of detailing the requirements incorrectly which would lead to an undesirable result. An iterative approach proved to be the right one, since as the project progressed changes to the requirements occurred due to a growing understanding of the business by the developers and a deeper understanding of the technical possibilities.
The project methodology adopted by the consortium was DSDM (Dynamic System Development Method), whose inherent features include a prototyping approach and other features that were adapted to meet the needs of the Descartes project.
The core principles used in DSDM are:
- active user involvement;
- empowerment of the core project team to enable quick and timely decisions;
- frequent deliverables in agreed timeboxes;
- iterative and incremental developments, including testing in each timebox;
- requirements are base-lined at a high level.
Overall this approach has worked well, with regular prototyping allowing corrections to be made and risks to be addressed (for instance, the good results from the Descartes Passenger Recovery System has allowed development to be focused more on this system). On the other hand initial difficulties experienced with the development of the Crew System (due to the complexity of how the flight crew are managed on the day) required a more lengthy development on this system.
The project was divided into three groups of activities.
- The first group was user oriented and mainly concerned with the definition of the problem and the organisational setting where the product was going to be used as well as serving as management of the project.
- The second group was concerned with how to solve the problem. It was centred around three prototypes each addressing the disruption management problem, each in more detail than the previous one. For each of these cases, optimisation models were developed which handled the integrated problem of aircraft, crew and passengers. To support the optimisation models (and to make sure that the cases map all relevan
Funding
Results
The result of the Descartes project is the development of a decision support tools designed to handle all sizes of disruption and able to match or better the decisions that the Operations Controllers make (in terms of speed and cost).
The Descartes system is made up of a set of sub systems (Dedicated Passenger Recovery Solver, Dedicated Aircraft Recovery Solver, Dedicated Crew Recovery Solver), which are able to solve the disruption from a single resource point of view. There are 3 solvers, aircraft, crew and passenger and each system has a set of parameters to which costs are associated. These parameters are configurable depending on the business strategy for managing disruption (e.g. if the strategy is to delay as opposed to cancel flights, the cancellation cost would be set as expensive, therefore the solution would be steered towards delaying flights instead). In this respect the system is flexible in how solutions are reached. The solutions are also structurally different in their make up, allowing choice to the controller as to the type of solution they would prefer to implement. The sub system solutions are then brought together through a messaging hub, the solutions are then compared and an integrated solution is produced.
In general, the consortium believes that the Descartes project has been a resounding success. The core objectives set out at the beginning of the project, in the main, have been achieved. There are prototypes of a system that, although needing further work to productionise, integrates aircraft, crew and passengers to give quality solutions to operational disruption situations.
Among the key successes are:
- Descartes added value to the business has been proven through a set of business experiments, whereby realistic scenarios have been written and run through both with a controller and the system. In all cases it has been seen that the system has been able to come up with solutions that are of similar or better quality, (with regard to cost) to those of the controllers and the amount of time taken to reach a solution through using the system has been radically reduced.
- Further funding from the Danish Technical Research Council awarded to DTU to further research disruption management for the wider transport industry, looking at the techniques used in Descartes and whether or not these would be applicable outside of the airline industry.
- Carmen Systems has grown considerably over the past three
Technical Implications
Future research is envisaged to address the following issues.
1) Disruption Management in other businesses.
The routing of vehicles, production planning in shipyards, and network planning in telecommunication constitute application areas, in which DTU (Danish Technical University) has already initiated work, mainly in terms of Ph.D projects and MSc projects. The initial experiences have been that both the methodology applied in Descartes (early user involvement, prototype development, and frequent user contact) and the particular combination of methods (heuristics, network models, and the combination of simulation and optimisation) have proven valuable. From a system development point of view, the experiences (with respect to user acceptance and implementation) in the development of an IT-based planning system based on mathematical methods seems particularly valuable. Currently, DTU is initiating a project involving decision support in the pharmaceutical industry and is applying for a research grant enabling the study of disruption management methods in other sectors of the transportation industry (road and sea transport).
2) Integrated planning systems.
The problem of designing and implementing a truly integrated disruption management system is very complicated and in Descartes, only initial steps have been taken. In the field of integrated planning systems, work has been done along similar lines (aircraft each with one crew) to allow "proof of concept", i.e. that plans taking into account several resources simultaneously are competitive with plans made sequentially (first aircraft, then crew). The planning builds on integrated mathematical models for both aircraft and crew and the solution methods include traditional mathematical programming techniques, which are feasible due to the relaxed timing constraints. However, making use of such methods in real-time for disruption management is not feasible. It is an open issue how these methods may be combined with heuristics in order to obtain a method, which makes use of the best of both worlds.
3) Simulation and optimisation.
In many real-life planning problems it is difficult to evaluate the expected effect of a proposed plan due to the fact that the benefit/cost of the plan cannot be expressed in a closed form, i.e. as a mathematical function of the data and parameters of the problem. In the context of Disr