Real-Time Local GP Model Learning

  • Duy Nguyen-Tuong
  • Matthias Seeger
  • Jan Peters
Part of the Studies in Computational Intelligence book series (SCI, volume 264)

Abstract

For many applications in robotics, accurate dynamics models are essential. However, in some applications, e.g., in model-based tracking control, precise dynamics models cannot be obtained analytically for sufficiently complex robot systems. In such cases, machine learning offers a promising alternative for approximating the robot dynamics using measured data. However, standard regression methods such as Gaussian process regression (GPR) suffer from high computational complexity which prevents their usage for large numbers of samples or online learning to date. In this paper, we propose an approximation to the standard GPR using local Gaussian processes models inspired by [Vijayakumar et al(2005)Vijayakumar, D’Souza, and Schaal, Snelson and Ghahramani(2007)]. Due to reduced computational cost, local Gaussian processes (LGP) can be applied for larger sample-sizes and online learning. Comparisons with other nonparametric regressions, e.g., standard GPR, support vector regression (SVR) and locally weighted projection regression (LWPR), show that LGP has high approximation accuracy while being sufficiently fast for real-time online learning.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [Chang and Lin(2001)]
    Chang, C.C., Lin, C.J.: LIBSVM: a library for support vector machines (2001), http://www.csie.ntu.edu.tw/~cjlin/libsvm
  2. [Craig(2004)]
    Craig, J.J.: Introduction to Robotics: Mechanics and Control, 3rd edn. Prentice Hall, Englewood Cliffs (2004)Google Scholar
  3. [Csato and Opper(2002)]
    Csato, L., Opper, M.: Sparse online gaussian processes. Neural Computation (2002)Google Scholar
  4. [Fumagalli et al(2009)Fumagalli, Gijsberts, Ivaldi, Jamone, Metta, Natale, Nori, and Sandini]
    Fumagalli, M., Gijsberts, A., Ivaldi, S., Jamone, L., Metta, G., Natale, L., Nori, F., Sandini, G.: Learning how to exploit proximal force sensing: a comparison approach. In: Sigaud, O., Peters, J. (eds.) From Motor Learning to Interaction Learning in Robots. SCI, vol. 264, pp. 149–169. Springer, Heidelberg (2010)Google Scholar
  5. [M.Seeger(2005)]
    MSeeger, Bayesian gaussian process models: Pac-bayesian generalisation error bounds and sparse approximations. PhD thesis, University of Edinburgh (2005)Google Scholar
  6. [M.Seeger(2007)]
    MSeeger. Low rank update for the cholesky decomposition. Tech. rep., University of California at Berkeley (2007), http://www.kyb.tuebingen.mpg.de/bs/people/seeger/
  7. [Nakanishi et al(2005)Nakanishi, Farrell, and Schaal]
    Nakanishi, J., Farrell, J.A., Schaal, S.: Composite adaptive control with locally weighted statistical learning. Neural Networks (2005)Google Scholar
  8. [Nguyen-Tuong et al(2008)Nguyen-Tuong, Peters, and Seeger]
    Nguyen-Tuong, D., Peters, J., Seeger, M.: Computed torque control with nonparametric regression models. In: Proceedings of the 2008 American Control Conference, ACC 2008 (2008)Google Scholar
  9. [Rasmussen and Williams(2006)]
    Rasmussen, C.E., Williams, C.K.: Gaussian Processes for Machine Learning. MIT-Press, Massachusetts Institute of Technology (2006)Google Scholar
  10. [Roberts et al(2009)Roberts, Zhang, and Tedrake]
    Roberts, J.W., Moret, L., Zhang, J., Tedrake, R.: Motor Learning at Intermediate Reynolds Number: Experiments with Policy Gradient on the Flapping Flight of a RigidWing. In: Sigaud, O., Peters, J. (eds.) From Motor Learning to Interaction Learning in Robots. SCI, vol. 264, pp. 293–309. Springer, Heidelberg (2010)Google Scholar
  11. [Schaal et al(2000)Schaal, Atkeson, and Vijayakumar]
    Schaal, S., Atkeson, C.G., Vijayakumar, S.: Real-time robot learning with locally weighted statistical learning. In: International Conference on Robotics and Automation (2000)Google Scholar
  12. [Schaal et al(2002)Schaal, Atkeson, and Vijayakumar]
    Schaal, S., Atkeson, C.G., Vijayakumar, S.: Scalable techniques from nonparameteric statistics for real-time robot learning. In: Applied Intelligence, pp. 49–60 (2002)Google Scholar
  13. [Schölkopf and Smola(2002)]
    Schölkopf, B., Smola, A.: Learning with Kernels: Support Vector Machines, Regularization, Optimization and Beyond. MIT Press, Cambridge (2002)Google Scholar
  14. [Seeger(2004)]
    Seeger, M.: Gaussian processes for machine learning. International Journal of Neural Systems (2004)Google Scholar
  15. [Seeger(2007)]
    Seeger, M.: LHOTSE: Toolbox for Adaptive Statistical Model (2007), http://www.kyb.tuebingen.mpg.de/bs/people/seeger/lhotse/
  16. [Snelson and Ghahramani(2007)]
    Snelson, E., Ghahramani, Z.: Local and global sparse gaussian process approximations. Artificial Intelligence and Statistics (2007)Google Scholar
  17. [Spong et al(2006)Spong, Hutchinson, and Vidyasagar]
    Spong, M.W., Hutchinson, S., Vidyasagar, M.: Robot Dynamics and Control. John Wiley and Sons, New York (2006)Google Scholar
  18. [Vijayakumar et al(2005)Vijayakumar, D’Souza, and Schaal]
    Vijayakumar, S., D’Souza, A., Schaal, S.: Incremental online learning in high dimensions. Neural Computation (2005)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Duy Nguyen-Tuong
    • 1
  • Matthias Seeger
    • 2
  • Jan Peters
    • 1
  1. 1.Max Planck Institute for Biological CyberneticsTübingen
  2. 2.Saarland UniversitySaarbrücken

Personalised recommendations