Federated Learning Improves Accuracy in Robotic Arm Inverse Dynamic Models

Strategic Insights -

Researchers from the University of Granada, in collaboration with Colombian and Spanish institutions, have developed an innovative approach to machine learning for robotic arms. Despite the challenges of accessing real-world data in robotics due to restrictions on sharing and limited availability, the team employed federated learning to train a model without centralizing data. This breakthrough has led to improved accuracy in their federated solution, with a 20% increase for the learned inverse dynamic model.

Key Takeaways:

  • The study, funded by the Consejeria de Transformacion Economica,Industria, Conocimiento y Universidades de la Junta de Andalucia, Colombian Ministry of Science, Technology, and Innovation, Spanish National - MICIU/AEI/10.13039/501100011033, and European Union (EU), found that access to real-world data in robotics is challenging due to privacy and intellectual property concerns.
  • The researchers proposed a solution that uses federated learning to train a model from distributed data, developing a robust robotic arm inverse dynamic model.
  • The custom aggregation method, integrating locally learned solutions from different workspaces, demonstrated improved accuracy by approximately 20% for the learned inverse dynamic model.
  • The investigation employed an informed federated learning approach to train a robotic arm inverse dynamic model, showing promising results for nonrigid robot identification.
  • The work has been peer-reviewed and published in Ieee Robotics and Automation Letters (2025;10(10):11022-11029).
  • The research team consisted of Gabriel Jimenez-Perera, University of Granada, Brayan Valencia-Vidal, Niceto R. Luque, Eduardo Ros, and Francisco Barranco.

Statistics:

  • 20% improvement in accuracy for the learned inverse dynamic model using federated learning.
  • Peer-reviewed research in Ieee Robotics and Automation Letters (2025 volume, issue 10).
  • Funded by several institutions, including:

+ Consejeria de Transformacion Economica,Industria, Conocimiento y Universidades de la Junta de Andalucia

+ Colombian Ministry of Science, Technology, and Innovation

+ Spanish National - MICIU/AEI/10.13039/501100011033

+ European Union (EU)

Sources:

  • Informed Federated Learning To Train a Robotic Arm Inverse Dynamic Model, Ieee Robotics and Automation Letters (2025;10(10):11022-11029)
  • NewsRx. Recent Research from University of Granada Highlight Findings in Machine Learning (Informed Federated Learning To Train a Robotic Arm Inverse Dynamic Model). Information Technology Newsweekly. October 21, 2025; p 618.