Abstract
This paper develops a data-driven iterative method for inverse optimal control when system dynamics are unknown. It leverages deep Koopman representations and parameterized objective features to jointly estimate dynamics and cost-function weights from observed states and inputs. Simulations validate the approach on nonlinear optimal-control settings.
Demo
Demo coming soon.
Citation
Liang, Zihao, Wenjian Hao, and Shaoshuai Mou. 2023. "A data-driven approach for inverse optimal control." Proceedings of the 62nd IEEE Conference on Decision and Control (CDC), 3632-3637.
@inproceedings{ZLiang2023datadriv,
author = {Z Liang, W Hao, S Mou},
year = {2023},
title = {A data-driven approach for inverse optimal control},
booktitle = {Proceedings of the 62nd IEEE Conference on Decision and Control (CDC), 3632-3637},
url = {https://ieeexplore.ieee.org/iel7/10383192/10383193/10383220.pdf}
}