Logo Leibniz Universität Hannover
Logo: Institute of Automatic Control
Logo Leibniz Universität Hannover
Logo: Institute of Automatic Control
  • Zielgruppen
  • Suche

Machine Learning

  • Multi-agent reinforcement learning
  • Learning of optimal motions/trajectories
  • Generalization of optimal trajectories
  • Task learning and -adaption

Non-linear Control Theory and Machine Learning

The research at IRT connects modern non-linear and optimal control of complex dynamical systems with methods from machine learning. Central results and new approaches for the systematic combination of optimal control of complex non-linear systems and learning and generalization were developed based on extending state-of-the-art machine learning methods. The results were successfully validated in highly complex robotic applications, allowing to find and apply almost optimal solutions for highly non-linear problems in real-time.



Höhn, O. (2008): Erkennung, Klassifikation und Vermeidung von Stürzen zweibeiniger Roboter, Norderstedt, Books On Demand. Dissertation, Leibniz Universität Hannover. more


Book Chapter

Haddadin, S., Weitschat, R., Huber, F., Özparpucu, M. C., Mansfeld, N., Albu-Schäffer, A. (2016): Optimal Control for Viscoelastic Robots and Its Generalization in Real-Time, Inaba, M., Corke, P. (Eds.): Robotics Research: The 16th International Symposium ISRR, Springer International Publishing, 131-148
DOI: 10.1007/978-3-319-28872-7_8


Li, Y. and Ganesh, G.; Jarrasse, N.; Haddadin, S.; Albu-Schäffer, A. & Burdet, E. (2018): Force, Impedance and Trajectory Learning for Contact Tooling and Haptic Identification, Accepted for IEEE Transactions on Robotics

Höhn, O. & Gerth, W. (2009): Probabilistic Balance Monitoring for Bipedal Robots, The International Journal of Robotics Research, vol. 28, no. 2, Feb, pp. 245-256.
DOI: 10.1177/0278364908095170


Conference Papers

Hu, Tingli; and Kühn, Johannes; and Ma'touq, Jumana & Haddadin, Sami (2018): Learning and Identification of human upper-limb muscle synergies in daily-life tasks with autoencoders, OTWorld Congress, Leipzig, Germany, 15.-18. May, more

Diaz Ledezma, Fernando & Haddadin, Sami (2017): First-Order-Principles-Based Constructive Network Topologies: An Application to Robot Inverse Dynamics, IEEE RAS International Conference on Humanoid Robots, Birmingham, UK

Golz, S., Osendorfer, Ch. & Haddadin, S. (2015): Using tactile sensation for learning contact knowledge: Discrimination collision from physical interaction, Accepted at: 2015 IEEE International Conference on Robotics and Automation
DOI: 10.1109/ICRA.2015.7139726


Tomic, T., Maier, M. & Haddadin, S. (2014): Learning Quadrocopter Maneuvers From Optimal Control and Generalizing in Real-time, International Conference on Robotics and Automation (ICRA), 2014, Hong Kong, China, May 31 - June 7, pp. 1747-1754
DOI: 10.1109/ICRA.2014.6907087


Haddadin, S., Weitschat, R., Huber, F., Özparpucu, M. C., Mansfeld, N. & Alin, A.-S. (2013): Optimal Control for Viscoelastic Robots and its Generalization in Real-Time, International Symposium on Robotics Research (ISRR).


Weitschat, R., Haddadin, S., Huber, F. & Albu-Schäffer, A. (2013): Dynamic optimality in real-time: A learning framework for near-optimal robot motions, Intelligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on, pp. 5636-5643.
DOI: 10.1109/IROS.2013.6697173


Höhn, O. & Gerth, W. (2008): Wahrscheinlichkeitsbasierte Sturzklassifikation von zweibeinigen Robotern, 42. Regelungstechnisches Kolloquium -- Kurzfassung der Beiträge, Feb., Boppard, pp. 39-40. more


Höhn, O., Schollmeyer, M. & Gerth, W. (2004): Sturzvermeidung von zweibeinigen Robotern durch reflexartige Reaktionen, In Holleczek, P. & Vogel-Heuser, B. (Ed.): Eingebettete Systeme. PEARL 2004, Informatik aktuell, Berlin Heidelberg, Springer, pp. 60-69. more
DOI: 10.1007/978-3-642-18594-6_7