Uncertainty Averse Pushing with Model Predictive Path Integral Control

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

Standard

Uncertainty Averse Pushing with Model Predictive Path Integral Control. / Arruda, Ermano; Jacob Mathew, Michael; Kopicki, Marek; Mistry, Michael; Azad, Morteza; Wyatt, Jeremy.

Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017). IEEE Computer Society Press, 2017. p. 497-502.

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

Harvard

Arruda, E, Jacob Mathew, M, Kopicki, M, Mistry, M, Azad, M & Wyatt, J 2017, Uncertainty Averse Pushing with Model Predictive Path Integral Control. in Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017). IEEE Computer Society Press, pp. 497-502, 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017), Birmingham, United Kingdom, 15/11/17. https://doi.org/10.1109/HUMANOIDS.2017.8246918

APA

Arruda, E., Jacob Mathew, M., Kopicki, M., Mistry, M., Azad, M., & Wyatt, J. (2017). Uncertainty Averse Pushing with Model Predictive Path Integral Control. In Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017) (pp. 497-502). IEEE Computer Society Press. https://doi.org/10.1109/HUMANOIDS.2017.8246918

Vancouver

Arruda E, Jacob Mathew M, Kopicki M, Mistry M, Azad M, Wyatt J. Uncertainty Averse Pushing with Model Predictive Path Integral Control. In Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017). IEEE Computer Society Press. 2017. p. 497-502 https://doi.org/10.1109/HUMANOIDS.2017.8246918

Author

Arruda, Ermano ; Jacob Mathew, Michael ; Kopicki, Marek ; Mistry, Michael ; Azad, Morteza ; Wyatt, Jeremy. / Uncertainty Averse Pushing with Model Predictive Path Integral Control. Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017). IEEE Computer Society Press, 2017. pp. 497-502

Bibtex

@inproceedings{42d2e0c665294948ba377fa6d092bb10,
title = "Uncertainty Averse Pushing with Model Predictive Path Integral Control",
abstract = "Planning robust robot manipulation requires good forward models that enable robust plans to be found. This work shows how to achieve this using a forward model learned from robot data to plan push manipulations. We explore learning methods (Gaussian Process Regression, and an Ensemble of Mixture Density Networks) that give estimates of the uncertainty in their predictions. These learned models are utilised by a model predictive path integral (MPPI) controller to plan how to push the box to a goal location. The planner avoids regions of high predictive uncertainty in the forward model. This includes both inherent uncertainty in dynamics, and meta uncertainty due to limited data. Thus, pushing tasks are completed in a robust fashion with respect to estimated uncertainty in the forward model and without the need of differentiable cost functions. We demonstrate the method on a real robot, and show that learning can outperform physics simulation. Using simulation, we also show the ability to plan uncertainty averse paths.",
keywords = "Cost function, Data models, Uncertainty, Predictive models, Planning, Robots, Trajectory",
author = "Ermano Arruda and {Jacob Mathew}, Michael and Marek Kopicki and Michael Mistry and Morteza Azad and Jeremy Wyatt",
year = "2017",
month = nov,
day = "15",
doi = "10.1109/HUMANOIDS.2017.8246918",
language = "English",
isbn = "978-1-5386-4678-6",
pages = "497--502",
booktitle = "Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017)",
publisher = "IEEE Computer Society Press",
note = "2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017) ; Conference date: 15-11-2017 Through 17-11-2017",

}

RIS

TY - GEN

T1 - Uncertainty Averse Pushing with Model Predictive Path Integral Control

AU - Arruda, Ermano

AU - Jacob Mathew, Michael

AU - Kopicki, Marek

AU - Mistry, Michael

AU - Azad, Morteza

AU - Wyatt, Jeremy

PY - 2017/11/15

Y1 - 2017/11/15

N2 - Planning robust robot manipulation requires good forward models that enable robust plans to be found. This work shows how to achieve this using a forward model learned from robot data to plan push manipulations. We explore learning methods (Gaussian Process Regression, and an Ensemble of Mixture Density Networks) that give estimates of the uncertainty in their predictions. These learned models are utilised by a model predictive path integral (MPPI) controller to plan how to push the box to a goal location. The planner avoids regions of high predictive uncertainty in the forward model. This includes both inherent uncertainty in dynamics, and meta uncertainty due to limited data. Thus, pushing tasks are completed in a robust fashion with respect to estimated uncertainty in the forward model and without the need of differentiable cost functions. We demonstrate the method on a real robot, and show that learning can outperform physics simulation. Using simulation, we also show the ability to plan uncertainty averse paths.

AB - Planning robust robot manipulation requires good forward models that enable robust plans to be found. This work shows how to achieve this using a forward model learned from robot data to plan push manipulations. We explore learning methods (Gaussian Process Regression, and an Ensemble of Mixture Density Networks) that give estimates of the uncertainty in their predictions. These learned models are utilised by a model predictive path integral (MPPI) controller to plan how to push the box to a goal location. The planner avoids regions of high predictive uncertainty in the forward model. This includes both inherent uncertainty in dynamics, and meta uncertainty due to limited data. Thus, pushing tasks are completed in a robust fashion with respect to estimated uncertainty in the forward model and without the need of differentiable cost functions. We demonstrate the method on a real robot, and show that learning can outperform physics simulation. Using simulation, we also show the ability to plan uncertainty averse paths.

KW - Cost function

KW - Data models

KW - Uncertainty

KW - Predictive models

KW - Planning

KW - Robots

KW - Trajectory

U2 - 10.1109/HUMANOIDS.2017.8246918

DO - 10.1109/HUMANOIDS.2017.8246918

M3 - Conference contribution

SN - 978-1-5386-4678-6

SP - 497

EP - 502

BT - Proceedings of 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017)

PB - IEEE Computer Society Press

T2 - 2017 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2017)

Y2 - 15 November 2017 through 17 November 2017

ER -