A First Principles Approach for Data-Efficient System Identification of Spring-Rod Systems via Differentiable Physics Engines

Kun Wang, Mridul Aanjaneya, Kostas Bekris
; Proceedings of the 2nd Conference on Learning for Dynamics and Control, PMLR 120:651-665, 2020.

Abstract

We propose a novel differentiable physics engine for system identification of complex spring-rod assemblies. Unlike black-box data-driven methods for learning the evolution of a dynamical system and its parameters, we modularize the design of our engine using a discrete form of the governing equations of motion, similar to a traditional physics engine. We further reduce the dimension from 3D to 1D for each module, which allows efficient learning of system parameters using linear regression. As a side benefit, the regression parameters correspond to physical quantities, such as spring stiffness or the mass of the rod, making the pipeline explainable. The approach significantly reduces the amount of training data required, and also avoids iterative identification of data sampling and model training. We compare the performance of the proposed engine with previous solutions, and demonstrate its efficacy on tensegrity systems, such as NASA’s icosahedron.

Cite this Paper


BibTeX
@InProceedings{pmlr-v120-wang20b, title = {A First Principles Approach for Data-Efficient System Identification of Spring-Rod Systems via Differentiable Physics Engines}, author = {Wang, Kun and Aanjaneya, Mridul and Bekris, Kostas}, booktitle = {Proceedings of the 2nd Conference on Learning for Dynamics and Control}, pages = {651--665}, year = {2020}, editor = {Alexandre M. Bayen and Ali Jadbabaie and George Pappas and Pablo A. Parrilo and Benjamin Recht and Claire Tomlin and Melanie Zeilinger}, volume = {120}, series = {Proceedings of Machine Learning Research}, address = {The Cloud}, month = {10--11 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v120/wang20b/wang20b.pdf}, url = {http://proceedings.mlr.press/v120/wang20b.html}, abstract = {We propose a novel differentiable physics engine for system identification of complex spring-rod assemblies. Unlike black-box data-driven methods for learning the evolution of a dynamical system and its parameters, we modularize the design of our engine using a discrete form of the governing equations of motion, similar to a traditional physics engine. We further reduce the dimension from 3D to 1D for each module, which allows efficient learning of system parameters using linear regression. As a side benefit, the regression parameters correspond to physical quantities, such as spring stiffness or the mass of the rod, making the pipeline explainable. The approach significantly reduces the amount of training data required, and also avoids iterative identification of data sampling and model training. We compare the performance of the proposed engine with previous solutions, and demonstrate its efficacy on tensegrity systems, such as NASA’s icosahedron.} }
Endnote
%0 Conference Paper %T A First Principles Approach for Data-Efficient System Identification of Spring-Rod Systems via Differentiable Physics Engines %A Kun Wang %A Mridul Aanjaneya %A Kostas Bekris %B Proceedings of the 2nd Conference on Learning for Dynamics and Control %C Proceedings of Machine Learning Research %D 2020 %E Alexandre M. Bayen %E Ali Jadbabaie %E George Pappas %E Pablo A. Parrilo %E Benjamin Recht %E Claire Tomlin %E Melanie Zeilinger %F pmlr-v120-wang20b %I PMLR %J Proceedings of Machine Learning Research %P 651--665 %U http://proceedings.mlr.press %V 120 %W PMLR %X We propose a novel differentiable physics engine for system identification of complex spring-rod assemblies. Unlike black-box data-driven methods for learning the evolution of a dynamical system and its parameters, we modularize the design of our engine using a discrete form of the governing equations of motion, similar to a traditional physics engine. We further reduce the dimension from 3D to 1D for each module, which allows efficient learning of system parameters using linear regression. As a side benefit, the regression parameters correspond to physical quantities, such as spring stiffness or the mass of the rod, making the pipeline explainable. The approach significantly reduces the amount of training data required, and also avoids iterative identification of data sampling and model training. We compare the performance of the proposed engine with previous solutions, and demonstrate its efficacy on tensegrity systems, such as NASA’s icosahedron.
APA
Wang, K., Aanjaneya, M. & Bekris, K.. (2020). A First Principles Approach for Data-Efficient System Identification of Spring-Rod Systems via Differentiable Physics Engines. Proceedings of the 2nd Conference on Learning for Dynamics and Control, in PMLR 120:651-665

Related Material