Soft Multicopter Control Using Neural Dynamics Identification

Yitong Deng, Yaorui Zhang, Xingzhe He, Shuqi Yang, Yunjin Tong, Michael Zhang, Daniel DiPietro, Bo Zhu
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:1773-1782, 2021.

Abstract

We propose a data-driven method to automatically generate feedback controllers for soft multicopters featuring deformable materials, non-conventional geometries, and asymmetric rotor layouts, to deliver compliant deformation and agile locomotion. Our approach coordinates two sub-systems: a physics-inspired network ensemble that simulates the soft drone dynamics and a custom LQR control loop enhanced by a novel online-relinearization scheme to control the neural dynamics. Harnessing the insights from deformation mechanics, we design a decomposed state formulation whose modularity and compactness facilitate the dynamics learning while its measurability readies it for real-world adaptation. Our method is painless to implement, and requires only conventional, low-cost gadgets for fabrication. In a high-fidelity simulation environment, we demonstrate the efficacy of our approach by controlling a variety of customized soft multicopters to perform hovering, target reaching, velocity tracking, and active deformation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-deng21a, title = {Soft Multicopter Control Using Neural Dynamics Identification}, author = {Deng, Yitong and Zhang, Yaorui and He, Xingzhe and Yang, Shuqi and Tong, Yunjin and Zhang, Michael and DiPietro, Daniel and Zhu, Bo}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {1773--1782}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/deng21a/deng21a.pdf}, url = {https://proceedings.mlr.press/v155/deng21a.html}, abstract = {We propose a data-driven method to automatically generate feedback controllers for soft multicopters featuring deformable materials, non-conventional geometries, and asymmetric rotor layouts, to deliver compliant deformation and agile locomotion. Our approach coordinates two sub-systems: a physics-inspired network ensemble that simulates the soft drone dynamics and a custom LQR control loop enhanced by a novel online-relinearization scheme to control the neural dynamics. Harnessing the insights from deformation mechanics, we design a decomposed state formulation whose modularity and compactness facilitate the dynamics learning while its measurability readies it for real-world adaptation. Our method is painless to implement, and requires only conventional, low-cost gadgets for fabrication. In a high-fidelity simulation environment, we demonstrate the efficacy of our approach by controlling a variety of customized soft multicopters to perform hovering, target reaching, velocity tracking, and active deformation.} }
Endnote
%0 Conference Paper %T Soft Multicopter Control Using Neural Dynamics Identification %A Yitong Deng %A Yaorui Zhang %A Xingzhe He %A Shuqi Yang %A Yunjin Tong %A Michael Zhang %A Daniel DiPietro %A Bo Zhu %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-deng21a %I PMLR %P 1773--1782 %U https://proceedings.mlr.press/v155/deng21a.html %V 155 %X We propose a data-driven method to automatically generate feedback controllers for soft multicopters featuring deformable materials, non-conventional geometries, and asymmetric rotor layouts, to deliver compliant deformation and agile locomotion. Our approach coordinates two sub-systems: a physics-inspired network ensemble that simulates the soft drone dynamics and a custom LQR control loop enhanced by a novel online-relinearization scheme to control the neural dynamics. Harnessing the insights from deformation mechanics, we design a decomposed state formulation whose modularity and compactness facilitate the dynamics learning while its measurability readies it for real-world adaptation. Our method is painless to implement, and requires only conventional, low-cost gadgets for fabrication. In a high-fidelity simulation environment, we demonstrate the efficacy of our approach by controlling a variety of customized soft multicopters to perform hovering, target reaching, velocity tracking, and active deformation.
APA
Deng, Y., Zhang, Y., He, X., Yang, S., Tong, Y., Zhang, M., DiPietro, D. & Zhu, B.. (2021). Soft Multicopter Control Using Neural Dynamics Identification. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:1773-1782 Available from https://proceedings.mlr.press/v155/deng21a.html.

Related Material