BROWSE
Detail
Conference Paper
시스템 매개변수를 활용한 Extended spKAE 기반 비선형 시스템 장기 예측
Cited - time in 
Share 
Share
- Issue Date
- 2026-06
- Citation
- 대한전자공학회 학술 대회 (하계) 2026, pp.1-5
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- This paper proposes extended spKAE (espKAE) for accurate long-horizon prediction of nonlinear systems with known system parameters. Koopman AutoEncoder (KAE) represents system dynamics in a latent space using a fixed Koopman operator, limiting its ability to capture dynamics that vary with states and system parameters. Though state- dependent polytopic KAE (spKAE) recomposes a Koopman operator from a predicted latent vector, it does not explicitly use system parameters. The proposed espKAE addresses this limitation by recomposing a Koopman operator using both the predicted latent vector and system parameters. We evaluate espKAE against KAE and spKAE on a 3- dimensional nonlinear system. The results show that espKAE outperforms KAE and spKAE in long-horizon prediction, implying the effectiveness of incorporating system parameters.
- KSP Keywords
- Capture dynamics, Koopman Operator, Latent space, System dynamics(SD), nonlinear systems, system parameters