Personalized Observation Normalization for Federated Reinforcement Learning in Simulation Environments with Heterogeneity

ArXi:2605.27385v1 Announce Type: cross Federated reinforcement learning (FedRL) enables multiple agents to collaboratively train a global policy without sharing raw data, making it ideal for privacy-sensitive applications. However, FedRL faces challenges in heterogeneous environments where differing state-transition dynamics lead to non-identical input distributions and imbalanced parameter updates during aggregation.