Thesis

My Master’s thesis focuses on GPS-denied navigation. I intend to publish it here once I complete it in the spring.

The ultimate objective is to develop a system that, given a reference image (from a previous year) and a target location, enables a drone to navigate to the target solely relying on its knowledge of its current position (with induced uncertainty due to noise) and the view of the surrounding area from the current year.

The primary goal is for the system to acquire the ability to navigate safely. If it encounters difficulties in reaching the target, it should have the capability to return to the starting point or another designated safe location.

This research project has significantly enriched my understanding of the research process, utilizing a GPU cluster and the current state-of-the-art in Reinforcement Learning.