Feudal Graph Reinforcement Learning

Tommaso Marzi, Arshjot Khehra, Andrea Cini +1

Graph-based representations and message-passing modular policies constitute prominent approaches to tackling composable control problems in reinforcement learning (RL). However, as…