Approximated Behavioral Metric-Based State Projection for Federated Reinforcement Learning

Abstract

Federated reinforcement learning (FRL) methods usually share the encrypted local state or policy information and help each client to learn from others while preserving everyone's privacy. In this work, we propose that sharing the approximated behavior metric-based state projection function is a promising way to enhance the performance of FRL and concurrently provides an effective protection of sensitive information. We introduce FedRAG, a FRL framework to learn a computationally practical projection function of states for each client and aggregating the parameters of projection functions at a central server. The FedRAG approach shares no sensitive task-specific information, yet provides information gain for each client. We conduct extensive experiments on the DeepMind Control Suite to demonstrate insightful results.

Cite

Text

Guo et al. "Approximated Behavioral Metric-Based State Projection for Federated Reinforcement Learning." International Joint Conference on Artificial Intelligence, 2025. doi:10.24963/IJCAI.2025/590

Markdown

[Guo et al. "Approximated Behavioral Metric-Based State Projection for Federated Reinforcement Learning." International Joint Conference on Artificial Intelligence, 2025.](https://mlanthology.org/ijcai/2025/guo2025ijcai-approximated/) doi:10.24963/IJCAI.2025/590

BibTeX

@inproceedings{guo2025ijcai-approximated,
  title     = {{Approximated Behavioral Metric-Based State Projection for Federated Reinforcement Learning}},
  author    = {Guo, Zengxia and An, Bohui and Lu, Zhongqi},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year      = {2025},
  pages     = {5298-5306},
  doi       = {10.24963/IJCAI.2025/590},
  url       = {https://mlanthology.org/ijcai/2025/guo2025ijcai-approximated/}
}