I am now a full professor at TU Dresden. For more information please visit my new website...

Roberto Calandra

Research Interests:

Reinforcement Learning
Gaussian Processes
Deep Learning

More Information

Curriculum Vitae Publications Google Citations DBLP Research Gate

Contact Information


Roberto Calandra joined Technische Universitaet Darmstadt (translates roughly as Darmstadt University of Technology) on January 2012 as a PhD student. There he will develop new "Probabilistic Methods in Learning for Control & Robotics" working with Jan Peters and Marc Deisenroth.

Previously, he achieved a Master degree in Machine Learning and Data Mining at the Aalto University (formerly known as Helsinki University of Technology or TKK) in Finland. There he had the opportunity to work on Gaussian Process as research assistant under the supervision of Aki Vehtari. Furthermore Roberto wrote his thesis An Exploration of Deep Belief Networks toward Adaptive Learning on the topic of Deep Learning under the supervision of Olli Simula with the collaboration of Federico Montesino Pouzols and Tapani Raiko.

Before, Roberto achieved his Bachelor in Computer Science engineering at the Università degli studi di Palermo in Italy. There he wrote his thesis Design and Build of a Robotics mobile platform [in Italian] under the supervision of Haris Dindo.

Research Interests

Roberto's main interest is what he calls "Artificial Stupidity", where opposite to the traditional Artifical Intelligence the emphasis is not on the "Intelligence" but on the adaptibility and continuous learning. For this purpose his interest includes but are not limited to: Machine Learning, Robotics and Self-organizing Systems.

The topic that he is currently developing is High-dimensional model-based Reinforcement Learning in the context of Robotics (which involves Gaussian Process modeling, Optimization, Reinforcement learning, and Deep Learning).


Key References

    •       Bib
      Calandra, R.; Seyfarth, A.; Peters, J.; Deisenroth, M. (2015). Bayesian Optimization for Learning Gaits under Uncertainty, Annals of Mathematics and Artificial Intelligence (AMAI).
    •     Bib
      Calandra, R. and Peters, J. and Deisenroth M.P. (2014). Pareto Front Modeling for Sensitivity Analysis in Multi-Objective Bayesian Optimization, NIPS Workshop on Bayesian Optimization 2014.
    •     Bib
      Calandra, R.; Ivaldi, S.; Deisenroth, M.; Peters, J. (2015). Learning Torque Control in Presence of Contacts using Tactile Sensing from Robot Skin, Proceedings of the International Conference on Humanoid Robots (HUMANOIDS).
    •     Bib
      Calandra, R.; Raiko, T.; Deisenroth, M.P.; Montesino Pouzols, F. (2012). Learning Deep Belief Networks from Non-Stationary Streams, International Conference on Artificial Neural Networks (ICANN).


  •     Bib
    Calandra, R.; Seyfarth, A.; Peters, J.; Deisenroth, M.P. (2014). An Experimental Comparison of Bayesian Optimization for Bipedal Locomotion, Proceedings of 2014 IEEE International Conference on Robotics and Automation (ICRA).

A full list of publications can be found on this page.

!!Software *Rprop Optimization Toolbox for MATLAB (No longer updated) *Rprop training for Neural Networks for MATLAB (No longer updated)

Teaching/Supervising Activities

I am currently looking for students for 1 Master Thesis and 1 Bachelor Thesis. If you are interested, contact me as soon as possible.


* Supervisor Master Thesis: Kai Steinert - Heteroscedastic Gaussian Processes for control (co-supervised by Filipe Veiga)

  • Supervisor Robot Learning Project Fall 2015: Treede F. and Konow P. and Bied M.- Control and Learning for a Bipedal Robot (jointly with Philipp Beckerle and Sasha Voloshina)
  • Supervisor Bachelor Thesis: Unverzagt Felix -
  • Supervisor Studienarbeit: Fritsche L. - Learning to walk on rough terrain


Locomotion (supervised by Katayon Radkhah) * Supervisor Robot Learning Project Spring 2014: Schnell, F. - Advanced Bayesian optimization models