FLAME: A Federated Learning Benchmark for Robotic Manipulation

2025-03-03

AI Summary

To address the poor scalability, weak environmental adaptability, and data privacy risks inherent in centralized paradigms for large-scale robotic manipulation policy training, this paper introduces FLAME—the first federated learning benchmark tailored for robotic manipulation. FLAME enables decentralized policy training across heterogeneous simulation environments (PyBullet and Gazebo), integrating federated learning (e.g., FedAvg), imitation learning, and distributed reinforcement learning. We construct a multi-environment federated dataset comprising over 160,000 expert demonstrations and provide a standardized training and evaluation framework. Experiments systematically validate, for the first time, the feasibility of mainstream federated algorithms on robotic manipulation tasks, revealing fundamental trade-offs among Non-IID data distributions, communication overhead, and policy generalization. FLAME establishes a foundation for privacy-preserving, scalable, and adaptable robotic learning.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Distributed Machine Learning & Federated Learning
• Natural Language Processing: Safety and Robustness

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Abstract

Recent progress in robotic manipulation has been fueled by large-scale datasets collected across diverse environments. Training robotic manipulation policies on these datasets is traditionally performed in a centralized manner, raising concerns regarding scalability, adaptability, and data privacy. While federated learning enables decentralized, privacy-preserving training, its application to robotic manipulation remains largely unexplored. We introduce FLAME (Federated Learning Across Manipulation Environments), the first benchmark designed for federated learning in robotic manipulation. FLAME consists of: (i) a set of large-scale datasets of over 160,000 expert demonstrations of multiple manipulation tasks, collected across a wide range of simulated environments; (ii) a training and evaluation framework for robotic policy learning in a federated setting. We evaluate standard federated learning algorithms in FLAME, showing their potential for distributed policy learning and highlighting key challenges. Our benchmark establishes a foundation for scalable, adaptive, and privacy-aware robotic learning.

Problem

• Addresses scalability and privacy in robotic manipulation training.
• Introduces FLAME for federated learning in robotic manipulation.
• Evaluates federated learning algorithms for distributed policy learning.

Innovation

• Federated learning benchmark for robotics
• Large-scale datasets from diverse environments
• Decentralized, privacy-preserving policy training