A Caltech Library Service

Learning Unstable Dynamics with One Minute of Data: A Differentiation-based Gaussian Process Approach

Jimenez Rodriguez, Ivan D. and Rosolia, Ugo and Ames, Aaron D. and Yue, Yisong (2021) Learning Unstable Dynamics with One Minute of Data: A Differentiation-based Gaussian Process Approach. . (Unpublished)

[img] PDF - Submitted Version
See Usage Policy.


Use this Persistent URL to link to this item:


We present a straightforward and efficient way to estimate dynamics models for unstable robotic systems. Specifically, we show how to exploit the differentiability of Gaussian processes to create a state-dependent linearized approximation of the true continuous dynamics. Our approach is compatible with most Gaussian process approaches for system identification, and can learn an accurate model using modest amounts of training data. We validate our approach by iteratively learning the system dynamics of an unstable system such as a 9-D segway (using only one minute of data) and we show that the resulting controller is robust to unmodelled dynamics and disturbances, while state-of-the-art control methods based on nominal models can fail under small perturbations.

Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription Paper
Rosolia, Ugo0000-0002-1682-0551
Ames, Aaron D.0000-0003-0848-3177
Yue, Yisong0000-0001-9127-1989
Additional Information:We would like to thank Andrew Singletary and Ellen Novoseller for their help in setting up the robotics and the mathematical derivations, respectively. This work is supported in part by NSF #1637598, NSF #1645832, NSF #1932091, and funding from AeroVironment and BMW.
Funding AgencyGrant Number
Record Number:CaltechAUTHORS:20210510-095024771
Persistent URL:
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:109029
Deposited By: Tony Diaz
Deposited On:10 May 2021 17:22
Last Modified:10 May 2021 17:22

Repository Staff Only: item control page