Optimizing Swarm Drone Coordinated Motion Using Graph Neural Networks

Abstract

Swarm robotics seeks to achieve coordinated collective behavior among multiple autonomous agents through local interactions and distributed decision-making. This thesis investigates the use of Graph Neural Networks (GNNs) for modeling and learning coordination mechanisms in drone swarms. SpeciĄcally, it explores how graph-based representations can effectively encode spatial relationships and how temporal modeling enhances the prediction of collective motion. A simulation framework was developed to generate multi-agent trajectories using interaction-based controllers grounded in physical models such as Gaussian and LennardŰJones potentials. These expert demonstrations were used to train a Graph Attention Net- work with Gated Recurrent Units (GAT–GRU), enabling the prediction of control forces from observed drone positions. Experimental evaluation across multiple swarm conĄgurations demonstrated that the trained model reproduces expert-like behavior with minimal performance loss while maintaining stable formation and goal convergence. The Ąndings highlight the potential of graph-based learning approaches to generalize swarm coordination policies across varying team sizes and environmental conditions.