Computed torque control with variable gains through Gaussian process regression

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Standard

Computed torque control with variable gains through Gaussian process regression. / Alberto, Nicolas Torres; Mistry, Michael; Stulp, Freek.

IEEE-RAS International Conference on Humanoid Robots. Vol. 2015-February IEEE Computer Society Press, 2015. p. 212-217 7041362.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Harvard

Alberto, NT, Mistry, M & Stulp, F 2015, Computed torque control with variable gains through Gaussian process regression. in IEEE-RAS International Conference on Humanoid Robots. vol. 2015-February, 7041362, IEEE Computer Society Press, pp. 212-217, 2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014, Madrid, Spain, 18/11/14. https://doi.org/10.1109/HUMANOIDS.2014.7041362

APA

Alberto, N. T., Mistry, M., & Stulp, F. (2015). Computed torque control with variable gains through Gaussian process regression. In IEEE-RAS International Conference on Humanoid Robots (Vol. 2015-February, pp. 212-217). [7041362] IEEE Computer Society Press. https://doi.org/10.1109/HUMANOIDS.2014.7041362

Vancouver

Alberto NT, Mistry M, Stulp F. Computed torque control with variable gains through Gaussian process regression. In IEEE-RAS International Conference on Humanoid Robots. Vol. 2015-February. IEEE Computer Society Press. 2015. p. 212-217. 7041362 https://doi.org/10.1109/HUMANOIDS.2014.7041362

Author

Alberto, Nicolas Torres ; Mistry, Michael ; Stulp, Freek. / Computed torque control with variable gains through Gaussian process regression. IEEE-RAS International Conference on Humanoid Robots. Vol. 2015-February IEEE Computer Society Press, 2015. pp. 212-217

Bibtex

@inproceedings{388e0a3f544f4e2abef3e23300dc77db,
title = "Computed torque control with variable gains through Gaussian process regression",
abstract = "In computed torque control, robot dynamics are predicted by dynamic models. This enables more compliant control, as the gains of the feedback term can be lowered, because the task of compensating for robot dynamics is delegated from the feedback to the feedforward term. Previous work has shown that Gaussian process regression is an effective method for learning computed torque control, by setting the feedforward torques to the mean of the Gaussian process. We extend this work by also exploiting the variance predicted by the Gaussian process, by lowering the gains if the variance is low. This enables an automatic adaptation of the gains to the uncertainty in the computed torque model, and leads to more compliant low-gain control as the robot learns more accurate models over time. On a simulated 7-DOF robot manipulator, we demonstrate how accurate tracking is achieved, despite the gains being lowered over time.",
author = "Alberto, {Nicolas Torres} and Michael Mistry and Freek Stulp",
year = "2015",
month = feb,
day = "12",
doi = "10.1109/HUMANOIDS.2014.7041362",
language = "English",
isbn = "9781479971749",
volume = "2015-February",
pages = "212--217",
booktitle = "IEEE-RAS International Conference on Humanoid Robots",
publisher = "IEEE Computer Society Press",
note = "2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014 ; Conference date: 18-11-2014 Through 20-11-2014",

}

RIS

TY - GEN

T1 - Computed torque control with variable gains through Gaussian process regression

AU - Alberto, Nicolas Torres

AU - Mistry, Michael

AU - Stulp, Freek

PY - 2015/2/12

Y1 - 2015/2/12

N2 - In computed torque control, robot dynamics are predicted by dynamic models. This enables more compliant control, as the gains of the feedback term can be lowered, because the task of compensating for robot dynamics is delegated from the feedback to the feedforward term. Previous work has shown that Gaussian process regression is an effective method for learning computed torque control, by setting the feedforward torques to the mean of the Gaussian process. We extend this work by also exploiting the variance predicted by the Gaussian process, by lowering the gains if the variance is low. This enables an automatic adaptation of the gains to the uncertainty in the computed torque model, and leads to more compliant low-gain control as the robot learns more accurate models over time. On a simulated 7-DOF robot manipulator, we demonstrate how accurate tracking is achieved, despite the gains being lowered over time.

AB - In computed torque control, robot dynamics are predicted by dynamic models. This enables more compliant control, as the gains of the feedback term can be lowered, because the task of compensating for robot dynamics is delegated from the feedback to the feedforward term. Previous work has shown that Gaussian process regression is an effective method for learning computed torque control, by setting the feedforward torques to the mean of the Gaussian process. We extend this work by also exploiting the variance predicted by the Gaussian process, by lowering the gains if the variance is low. This enables an automatic adaptation of the gains to the uncertainty in the computed torque model, and leads to more compliant low-gain control as the robot learns more accurate models over time. On a simulated 7-DOF robot manipulator, we demonstrate how accurate tracking is achieved, despite the gains being lowered over time.

UR - http://www.scopus.com/inward/record.url?scp=84945192193&partnerID=8YFLogxK

U2 - 10.1109/HUMANOIDS.2014.7041362

DO - 10.1109/HUMANOIDS.2014.7041362

M3 - Conference contribution

AN - SCOPUS:84945192193

SN - 9781479971749

VL - 2015-February

SP - 212

EP - 217

BT - IEEE-RAS International Conference on Humanoid Robots

PB - IEEE Computer Society Press

T2 - 2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014

Y2 - 18 November 2014 through 20 November 2014

ER -