RMIX: Learning Risk-Sensitive Policies for Cooperative Reinforcement Learning Agents
arXiv:2102.08159
Abstract
Current value-based multi-agent reinforcement learning methods optimize individual Q values to guide individuals' behaviours via centralized training with decentralized execution (CTDE). However, such expected, i.e., risk-neutral, Q value is not sufficient even with CTDE due to the randomness of rewards and the uncertainty in environments, which causes the failure of these methods to train coordinating agents in complex environments. To address these issues, we propose RMIX, a novel cooperative MARL method with the Conditional Value at Risk (CVaR) measure over the learned distributions of individuals' Q values. Specifically, we first learn the return distributions of individuals to analytically calculate CVaR for decentralized execution. Then, to handle the temporal nature of the stochastic outcomes during executions, we propose a dynamic risk level predictor for risk level tuning. Finally, we optimize the CVaR policies with CVaR values used to estimate the target in TD error during centralized training and the CVaR values are used as auxiliary local rewards to update the local distribution via Quantile Regression loss. Empirically, we show that our method significantly outperforms state-of-the-art methods on challenging StarCraft II tasks, demonstrating enhanced coordination and improved sample efficiency.
ICLR 2021 submission version: https://openreview.net/forum?id=1EVb8XRBDNr
References in corpus (9)
- Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
- StarCraft II: A New Challenge for Reinforcement Learning
- QTRAN: Learning to Factorize with Transformation for Cooperative Multi-Agent Reinforcement Learning
- Distributional Reinforcement Learning with Quantile Regression
- MAVEN: Multi-Agent Variational Exploration
- Algorithms for CVaR Optimization in MDPs
- Policy Gradient for Coherent Risk Measures
- Randomized Ensembled Double Q-Learning: Learning Fast Without a Model
- Cautious Reinforcement Learning via Distributional Risk in the Dual Domain