Sequential Decision-Making at Airbus: The Beluga Challenge2024HT
|
|
Course plan
Context
During the course, we will work towards submitting solvers to the Beluga
Challenge, which is a competition run by Airbus that starts in October.
More information about the competition can be found here:
https://tuples.ai/competition-challenge/
Recommended for
Advanced Master and PhD students
Goals
Develop solvers and submit them to the Beluga Challenge.
Prerequisites
Good theoretical and practical knowledge in some form of sequential
decision-making are recommended.
There are several tracks, allowing to approach the problem from various angles,
from model-based symbolic planning or reinforcement learning in the scalability
tracks, to logic-based reasoning and natural language generation in the
explainability track. Having some expertise in at least one of these fields is
highly recommended.
Organization
We will kick off the course in October, depending on when more details about
the competition are released. In a first meeting we will discuss the different
tracks of the competition and possible approaches to tackle these tracks.
Participants will form groups, jointly working on a solver that will be
submitted to the competition.
In the main phase, November-January, we will have regular meetings during which
we brainstorm approaches that could be promising, discuss issues that arise
during the implementation, as well as suggest improvements that could further
enhance the performance.
Every group will have a short presentation in which they explain their idea,
and a longer presentation towards the end to present their competition
submission.
Examination
Project work in groups and presentation of your approach.
Examiner
Daniel Gnad
Credits
9
Page responsible: Director of Graduate Studies