Learning Impedance Control for Contact-Rich Robotic Manipulation TU Darmstadt, Germany April 2025 - Ongoing Collaborators: Reza Nazmara (University of Porto), Alap Kshirsagar (TU Darmstadt), Cristiana de Farias (TU Darmstadt), Davide Tateo (Lund University), Jan Peters (TU Darmstadt), and A. Pedro Aguiar (University of Porto)

This project develops learning-based impedance control methods for robots operating in uncertain, contact-rich environments. By combining robust and adaptive control theory with reinforcement learning techniques, the system enables robotic manipulators to regulate force, stiffness, and motion dynamically during physical interaction tasks. The research focuses on improving robustness, safety, and precision in scenarios such as object transport, grasping, and unexpected collisions where accurate modeling of contact dynamics is difficult. Through simulation and real-world experiments, the project investigates how data-driven policies can enhance compliant manipulation performance while maintaining stability and generalization across varying tasks and environments.

Publications

Nazmara, R., Kshirsagar, A., Peters, J., & Aguiar, A.P., Robust Adaptive Backstepping Impedance Control of Robots in Unknown Environments, Mechatronics, and IFAC World Congress (IFAC 2026), 2026