Electrical Engineering and Computer Science

We propose a novel approach to control a swarm of drones. Leveraging Graph Neural Networks (GNNs), our approach aims to improve the robustness of the drone swarm system, making the swarm accomplish tasks in adverse and disturbing conditions. In our project, one example of such harsh conditions can be when one or more agents are biased or compromised. Related works exist leveraging GNNs to decentralize the global controllers and bring many benefits to the swarm control system for drones. However, whether applying GNNs can improve the robustness still needs to be explored. Therefore, our objective is to investigate this problem and verify if using GNNs can enhance the robustness of the systems.  Thesis advisor:  Professor Mohammad Al Faruque.