Deep Koopman learning of nonlinear time-varying systems

January 2024 · W Hao, B Huang, W Pan, D Wu, S Mou · Automatica 2024


Abstract

This paper presents a data-driven approach that approximates nonlinear time-varying dynamics by linear time-varying Koopman models learned with deep neural networks. It provides approximation-error analysis between original and lifted dynamics and demonstrates small prediction errors even under rapid system changes. Quadcopter simulations further show strong computational efficiency for control-oriented modeling.


Demo

Demo coming soon.


Citation

Hao, Wenjian, Bowen Huang, Wei Pan, Di Wu, and Shaoshuai Mou. 2024. "Deep Koopman learning of nonlinear time-varying systems." Automatica 159:111372.

@article{WHao2024deepkoop,
  author = {W Hao, B Huang, W Pan, D Wu, S Mou},
  year = {2024},
  title = {Deep Koopman learning of nonlinear time-varying systems},
  journal = {Automatica 159:111372},
  url = {https://www.sciencedirect.com/science/article/pii/S0005109823005381}
}