English

Federated Reinforcement Learning with Environment Heterogeneity

Machine Learning 2022-04-07 v1 Machine Learning

Abstract

We study a Federated Reinforcement Learning (FedRL) problem in which nn agents collaboratively learn a single policy without sharing the trajectories they collected during agent-environment interaction. We stress the constraint of environment heterogeneity, which means nn environments corresponding to these nn agents have different state transitions. To obtain a value function or a policy function which optimizes the overall performance in all environments, we propose two federated RL algorithms, \texttt{QAvg} and \texttt{PAvg}. We theoretically prove that these algorithms converge to suboptimal solutions, while such suboptimality depends on how heterogeneous these nn environments are. Moreover, we propose a heuristic that achieves personalization by embedding the nn environments into nn vectors. The personalization heuristic not only improves the training but also allows for better generalization to new environments.

Keywords

Cite

@article{arxiv.2204.02634,
  title  = {Federated Reinforcement Learning with Environment Heterogeneity},
  author = {Hao Jin and Yang Peng and Wenhao Yang and Shusen Wang and Zhihua Zhang},
  journal= {arXiv preprint arXiv:2204.02634},
  year   = {2022}
}

Comments

Artificial Intelligence and Statistics 2022

R2 v1 2026-06-24T10:39:27.416Z