Asem Khattab

Asem Khattab

Active student


Master's Thesis Project: Towards an Interactive Drone, a Reinforcement Learning Approach

An essential requirement for a successful employment of drones in industrial applications is the ability to correctly deal with physical contact with the environment. The traditional approach consists of applying what is known as impedance control (force control and interaction control are also often used terms).

The robot (the drone in our case) then behaves as an impedance in the environment with mass, damper and spring. The problem is: the selection of good impedance parameters ensuring the desired behavior is not an easy straightforward task. It requires either an accurate model of the environment where the contact will occur or a human expert tuning the parameters simply by trial and error.

The aim of this thesis is to investigate different approaches of improving the autonomy of interaction control in drones relying on reinforcement learning strategies.