EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2040. Simulated Annealing to Solve the Integrated Airline Fleet and Crew Recovery Problem
Invited abstract in session TC-54: Disruption management and recovery, stream Public Transport Optimization.
Tuesday, 12:30-14:00Room: S01 (building: 101)
Authors (first author is the speaker)
| 1. | Philip de Bruin
|
| Information and Computing Sciences, Utrecht University | |
| 2. | Marjan van den Akker
|
| Information and Computing Sciences, Utrecht University | |
| 3. | Kunal Kumar
|
| KLM Royal Dutch Airlines | |
| 4. | Lisanne Heuseveldt
|
| Information and Computing Sciences, Utrecht University | |
| 5. | Marc Paelinck
|
| KLM Royal Dutch Airlines |
Abstract
Airline operations are prone to delays and disruptions, since the schedules are generally tight and depend on a lot of resources. Since the COVID pandemic, most airlines also have shortages of these resources, making the problem worse. In case of disruptions, such as aircraft breakdowns or crew sickness, the schedule needs to be adjusted. This means either changing resource assignments or delaying or cancelling flights. Such adjustments need to resolve a disruption while minimizing overall costs and impact on passengers. For practical use, this needs to be done as close to real-time as possible.
We focus on both the aircraft and crew schedules. Resolving disruptions for these can be done sequentially, but decisions in one problem may lead to more conflicts in the other. Thus, we take an integrated approach. Such an approach comes with computational challenges, especially because of the runtime required for practical use.
For this, we develop a local search algorithm based on simulated annealing. Where, in order to not spend time on making changes in unaffected parts of the schedule, we test different neighbour generation and selection approaches that steer the local search into resolving these disruptions more directly. The work is done in collaboration with and with data from KLM Royal Dutch Airlines.
We show that our approach is very fast, achieving runtimes well below a minute. Furthermore, we also show that the disruptions are resolved in a cost-efficient manner.
Keywords
- Airline Applications
- Metaheuristics
Status: accepted
Back to the list of papers