Our work on SFs for MARL was accepted to AAMAS 2022!

Our paper “Disentangling Successor Features for Coordination in Multi-agent Reinforcement Learning” was accepted for publication as a full paper and oral presentation at AAMAS 2022! We show that our approach for multi-agent reinforcement learning (MARL), named DISentangling Successor features for Coordination (DISSC), improves performance and training time relative to existing methods in a variety of environments including the Starcraft Multi-agent Challenge (SMAC). Our results suggest that successor features (SFs) offers a promising approach to coordination in MARL, based on its ability to separate environment dynamics from reward structure.

Check out the talk and preprint!