Celebrating Diversity in Shared Multi-Agent Reinforcement Learning

By Chenghao Li, Chengjie Wu, Tonghan Wang, Jun Yang, Qianchuan Zhao, Chongjie Zhang. 2021

208 citations recorded by the source corpus.

Recently, deep multi-agent reinforcement learning (MARL) has shown the promise to solve complex cooperative tasks. Its success is partly because of parameter sharing among agents. However, such sharing may lead agents to behave similarly and limit their coordination capacity. In this paper, we aim to introduce diversity in both optimization and representation of shared multi-agent reinforcement learning. Specifically, we propose an information-theoretical regularization to maximize the mutual information between agents' identities and their trajectories, encouraging extensive exploration and diverse individualized behaviors. In representation, we incorporate agent-specific modules in the shared neural network architecture, which are regularized by L1-norm to promote learning sharing among agents while keeping necessary diversity. Empirical results show that our method achieves state-of-the-art performance on Google Research Football and super hard StarCraft II micromanagement tasks.

Subfields: Artificial Intelligence, Computer Science Applications, Economics and Econometrics, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yu Wang
2. Chao Yu
3. Alexandre M. Bayen
4. Eugene Vinitsky
5. Gao Huang
6. Akash Velu
7. Chenghao Li
8. Yiqin Yang

In Computer Science Applications

1. Yi Wu

In Economics and Econometrics

1. Eugene Vinitsky
2. Alexandre M. Bayen
3. Akash Velu
4. Chao Yu
5. Yi Wu
6. Yu Wang

In Reinforcement Learning

1. Yu Wang
2. Alexandre M. Bayen
3. Chao Yu
4. Eugene Vinitsky
5. Gao Huang
6. Akash Velu
7. Chenghao Li
8. Yiqin Yang

Citations in the corpus

• The Surprising Effectiveness of PPO in Cooperative Multi-Agent Games (2021, methodology): “• GRF : QMix [27] and SOTA methods including CDS [20] and TiKick [16].”
• TiKick: Towards Playing Multi-agent Football Full Games from Single-agent Demonstrations (2021, neutral): “In this experiment, we use QMIX (Rashid et al., 2018), MADDPG (Lowe et al., 2017), CDS (Li et al., 2021) and MAPPO (Yu et al., 2021) as the baselines.”
• Believe What You See: Implicit Constraint Approach for Offline Multi-Agent Reinforcement Learning (2021, methodology): “The current offline algorithms are unsuccessful in multi-agent tasks even though they adopt the modern value-decomposition structure [26, 48, 25].”