TY - GEN

T1 - Efficient Online Inference and Learning in Partially Known Nonlinear State-Space Models by Learning Expressive Degrees of Freedom Offline

AU - Ewering, Jan-Hendrik

AU - Volkmann, Björn

AU - Ehlers, Simon Friedrich Gerhard

AU - Seel, Thomas

AU - Meindl, Michael Bernhard

N1 - Publisher Copyright: © 2024 IEEE.

PY - 2024/12/16

Y1 - 2024/12/16

N2 - Intelligent real-world systems critically depend on expressive information about their system state and changing operation conditions, e.g., due to variation in temperature, location, wear, or aging. To provide this information, online inference and learning attempts to perform state estimation and (partial) system identification simultaneously. Current works combine tailored estimation schemes with flexible learningbased models but suffer from convergence problems and computational complexity due to many degrees of freedom in the inference problem (i. e., parameters to determine). To resolve these issues, we propose a procedure for data-driven offline conditioning of a highly flexible Gaussian Process (GP) formulation such that online learning is restricted to a subspace, spanned by expressive basis functions. Due to the simplicity of the transformed problem, a standard particle filter can be employed for Bayesian inference. In contrast to most existing works, the proposed method enables online learning of target functions that are nested nonlinearly inside a first-principles model. Moreover, we provide a theoretical quantification of the error, introduced by restricting learning to a subspace. A Monte-Carlo simulation study with a nonlinear battery model shows that the proposed approach enables rapid convergence with significantly fewer particles compared to a baseline and a state-of-the-art method.

AB - Intelligent real-world systems critically depend on expressive information about their system state and changing operation conditions, e.g., due to variation in temperature, location, wear, or aging. To provide this information, online inference and learning attempts to perform state estimation and (partial) system identification simultaneously. Current works combine tailored estimation schemes with flexible learningbased models but suffer from convergence problems and computational complexity due to many degrees of freedom in the inference problem (i. e., parameters to determine). To resolve these issues, we propose a procedure for data-driven offline conditioning of a highly flexible Gaussian Process (GP) formulation such that online learning is restricted to a subspace, spanned by expressive basis functions. Due to the simplicity of the transformed problem, a standard particle filter can be employed for Bayesian inference. In contrast to most existing works, the proposed method enables online learning of target functions that are nested nonlinearly inside a first-principles model. Moreover, we provide a theoretical quantification of the error, introduced by restricting learning to a subspace. A Monte-Carlo simulation study with a nonlinear battery model shows that the proposed approach enables rapid convergence with significantly fewer particles compared to a baseline and a state-of-the-art method.

UR - http://www.scopus.com/inward/record.url?scp=86000505657&partnerID=8YFLogxK

U2 - 10.1109/CDC56724.2024.10886241

DO - 10.1109/CDC56724.2024.10886241

M3 - Conference contribution

AN - SCOPUS:86000505657

SN - 979-8-3503-1634-6

T3 - Proceedings of the IEEE Conference on Decision and Control

SP - 4157

EP - 4164

BT - 2024 IEEE 63rd Conference on Decision and Control, CDC 2024

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 63rd IEEE Conference on Decision and Control, CDC 2024

Y2 - 16 December 2024 through 19 December 2024

ER -