The project is created in the THY Travel Hackhathon 2023 as the proof of concept that suggests reinforcement learning can be a good choice in solving the problem of airline scheduling.
Airline companies how to deal with management of hundreds of planes and thousands of passengers each day. In order to maximize the profit planes should always be full. This makes the very hard optimization problem.
Reinforcement learning has already proved its ability in dealing with optimization problem. Airline scheduling must be flexible since the numbers of passengers desiring to travel is affected by many factor. Being a model free algorithm, RL can handle different situations. Its negative side is that it requires a robust simulation environment.
Ahmet Furkan Akıncı - @ahmetf1
Erkam Kavak - @erkamkavak
Mehmet Batuhan Çok
Ahmet Selim Gül
- Collecting data from the API`s served by Turkish Airlines.
- Creation of a airline and airplane simulation with possible passenger counts.
- Implementing Reinforcement Learning algorithm on the simulation environment.
- Tuning parameters for better results.