Federated Distributional Reinforcement Learning with Distributional Critic Regularization
This work addresses safety-critical settings in federated reinforcement learning by preserving distributional information, though it is incremental as it builds on existing distributional RL and federated learning methods.
The paper tackled the problem of statistical multimodality and tail behavior being obscured in federated reinforcement learning by proposing federated distributional reinforcement learning with a distributional critic regularization method, resulting in reduced mean-smearing, improved safety proxies like lower catastrophe rates, and lower critic/policy drift compared to baselines.
Federated reinforcement learning typically aggregates value functions or policies by parameter averaging, which emphasizes expected return and can obscure statistical multimodality and tail behavior that matter in safety-critical settings. We formalize federated distributional reinforcement learning (FedDistRL), where clients parametrize quantile value function critics and federate these networks only. We also propose TR-FedDistRL, which builds a per client, risk-aware Wasserstein barycenter over a temporal buffer. This local barycenter provides a reference region to constrain the parameter averaged critic, ensuring necessary distributional information is not averaged out during the federation process. The distributional trust region is implemented as a shrink-squash step around this reference. Under fixed-policy evaluation, the feasibility map is nonexpansive and the update is contractive in a probe-set Wasserstein metric under evaluation. Experiments on a bandit, multi-agent gridworld, and continuous highway environment show reduced mean-smearing, improved safety proxies (catastrophe/accident rate), and lower critic/policy drift versus mean-oriented and non-federated baselines.