Informed Federated Learning to Train a Robotic Arm Inverse Dynamic Model
Gabriel Jimenez-Perera, Brayan Valencia-Vidal, Niceto R. Luque, Eduardo Ros, Francisco Barranco
AI summary
Problem
Training data-driven inverse dynamic models for collaborative robots requires large, diverse datasets, but industrial data privacy and intellectual property restrictions prevent centralized data sharing.
Approach
The authors apply federated learning to train a recurrent neural network inverse dynamic model across multiple robotic clients, introducing a custom aggregation method that uses 3D workspace bounding cubes as spatial metadata to weight client updates.
Key results
- Successful federated training of a cobot inverse dynamic model without raw data sharing
- Public release of a 1.8M-sample dataset across six workspaces on a Baxter arm
- FedCubeNQ aggregation reduces accuracy errors by approximately 20% over standard federated averaging
- Validated generalization across varying client counts and diverse motion trajectories
Why it matters
Enables manufacturers and researchers to collaboratively train complex robot dynamics models while preserving data privacy and intellectual property rights.
Abstract
Access to real-world data in robotics domains is often challenging due to restrictions on data sharing and limited availability. Although privacy and intellectual property concerns are the main barriers, ensuring data access is crucial for ad- vancing data-driven models. Specifically, machine-learning-based inverse dynamic models show promising results for nonrigid robot identification, but the data used to train them are often kept private due to intellectual property protections. Federated learning proposes a methodology to access such data without centralizing them in a single repository, thus avoiding intellectual property limitations. We propose a solution that uses federated learning to train a model from distributed data to develop a robust robotic arm inverse dynamic model. Our approach demonstrates the feasibility of using a machine learning method in which local robots train on their own data while collaborating without sharing raw information. Furthermore, we propose a novel custom aggregation method that integrates locally learned solutions from different workspaces into a single global model without requiring raw data sharing. This method improves accuracy in our federated solution by approximately 20% for the learned inverse dynamic model.