Duy Nguyen-Tuong has been pursuing his Ph.D. since 2007 at the Max-Planck Institute for Biological Cybernetics in the department of Bernhard Schölkopf under the supervision of Jan Peters. Before doing so, he studied control and automation engineering at the University of Stuttgart and the National University of Singapore.

His main research interest is the application of machine learning techniques in control and robotics. Currently, he is working on fast regression methods for online model learning in real-time which can be used, for example, in robot model-based control. This research objective is based on the finding that accurate analytical models, e.g., rigid body models, can not be obtained for sufficiently complex robot systems due to many unknown nonlinearities. In such cases, machine learning presents an appealing alternative for approximating those models from measured data. Learning from data will include all robot's nonlinearities resulting in a more accurate model and, thus, a better control performance. However, the main drawback of machine learning techniques is the excessive computational complexity preventing a straightforward application in robotics. Furthermore, in order to take full advantage of a learning approach, online learning is an absolute necessity as it allows the model adaption due to changes in the robot dynamics, load or the actuators. For most real-time applications, online model learning poses a difficult regression problem due to three constraints, i.e., firstly, the learning and prediction process should be very fast (e.g., learning needs to take place at a speed of 20-200Hz and prediction at 200Hz to a 1000Hz). Secondly, the learning system needs to be capable at dealing with large amounts of data (i.e., with data arriving at 200Hz, less than ten minutes of runtime will result in more than a million data points). And, thirdly, the data arrives as a continuous stream, thus, the model has to be continuously adapted to new training examples over time.

Duy Nguyen-Tuong and his collaborators attempt to tackle these problems by investigating and developing regression techniques capable for real-time online model learning. The ultimate aim is to establish machine learning techniques in model-based robot control. He co-organized a NIPS 2009 workshop on Probabilistic Approaches for Robotics and Control. More details of his research activities can be found at Research > Model Learning.

Research Interests: Robot Control, Dynamics, Machine Learning, Kernel Methods.
Biographical Information: Please see his curriculum vitae.
Publications: For the complete list of his publication, see here. Key references are shown below.
Collaborators: Stefan Schaal (USC), Bernhard Schölkopf and Jan Peters

Key References

1. Nguyen-Tuong, D.; Seeger, M.; Peters, J. (2009). Local Gaussian Process Regression for Real Time Online Model Learning and Control, Advances in Neural Information Processing Systems 22 (NIPS 2008), Cambridge, MA: MIT Press. [PDF]
   
   

2. Nguyen-Tuong, D.; Peters, J.; Seeger, M.; Schoelkopf, B. (2008). Computed Torque Control with Nonparametric Regressions Techniques, American Control Conference. [PDF]
   
   

3. Nguyen-Tuong, D.; Seeger, M.; Peters, J. (2009). Model Learning with Local Gaussian Process Regression, Advanced Robotics, 23, 15, pp.2015-2034.
   
   

4. Nguyen-Tuong, D.; Peters, J. (2010). Using Model Knowledge for Learning Inverse Dynamics, IEEE International Conference on Robotics and Automation.
   
   

Contact Information

Mail: Duy Nguyen-Tuong, Spemannstr. 38, 72076 Tuebingen, Germany
Phone: +49-7071-601-558
Fax: +49-7071-601-552
Email: duy@tuebingen.mpg.de