Abstract
VDN and QMIX are two popular value-based algorithms for cooperative MARL that learn a centralized action value function as a monotonic mixing of per-agent utilities. While this enables easy decentralization of the learned policy, the restricted joint action value function can prevent them from solving tasks that require significant coordination between agents at a given timestep. We show that this problem can be overcome by improving the joint exploration of all agents during training. Specifically, we propose a novel MARL approach called Universal Value Exploration (UneVEn) that learns a set of related tasks simultaneously with a linear decomposition of universal successor features. With the policies of already solved related tasks, the joint exploration process of all agents can be improved to help them achieve better coordination. Empirical results on a set of exploration games, challenging cooperative predator-prey tasks requiring significant coordination among agents, and StarCraft II micromanagement benchmarks show that UneVEn can solve tasks where other state-of-the-art MARL methods fail.
Original language | English |
---|---|
Title of host publication | Proceedings of the International Conference on Machine Learning (ICML) |
Editors | Marina Meila, Tong Zhang |
Pages | 3930-3941 |
Number of pages | 12 |
Volume | 139 |
Publication status | Published - 2021 |
Event | International Conference on Machine Learning: 2021 - Duration: 18 Jul 2021 → 24 Jul 2021 Conference number: 38th https://icml.cc/Conferences/2021 |
Publication series
Name | Proceedings of Machine Learning Research |
---|---|
Volume | PMLR 139 |
ISSN (Electronic) | 2640-3498 |
Conference
Conference | International Conference on Machine Learning |
---|---|
Abbreviated title | ICML |
Period | 18/07/21 → 24/07/21 |
Internet address |