Subject

A flight scheduler is a program that collects and processes data from multiple sources to plan flight paths for Unmanned Aerial Vehicles (UAVs). Current versions of flight schedules are not able to deal with the dynamism required for UAV flight planning.
The leveraging of a deep reinforcement learning (RL) based flight plan and scheduler, can help overcome limitations that exist with more traditional schedulers. These schedulers should be able to provide and update flight plans in a quicker and more efficient way, all in automated way.

Kind of work

A key component of the scheduler is the routing algorithm, which defines the optimal route that the UAV should follow to go from one point to another. These routing algorithms typically require data from multiple sources, including, but not limited to, power consumption, weather data and crowd data.
Live data needs to be integrated into the algorithm that produces the routes. A version of the routing algorithm exists, but the data that it has used in the past was historical data and the concept was tested to a limited extent. This thesis will focus more towards the inclusion of live data, such as GIPOD for event data or Proximus crowd data for identification of assemblies of people, to create a path for the drone at any given moment. Another type of data is live traffic, which would show other aerial vehicles in a certain region.
Through the presence of the data, it would also be possible to predict aerial and ground activity in the future.

Framework of the Thesis

This thesis is in collaboration with the company Helicus.

Number of Students

1

Expected Student Profile

Good knowledge of Machine Learning, AI and data processing. Good programming skills in Python (particularly PyTorch)