GNN-MARL Implementation

A GNN combined with MARL for coordinating drone swarms. The project is an implementation of QMIX-GNN and introduces some new features such as anomaly detection.

Source

A graph neural network (GNN) combined with multi-agent reinforcement learning (MARL) for coordinating drone swarms. I extended the QMIX-GNN Paper by adding features such as dynamic graph construction, attention mechanisms, and improved reward shaping to enhance the coordination and efficiency of drone swarms in complex environments.

Built with

PythonPytorchFlask