-
Space-Filling Regularization for Robust and Interpretable Nonlinear State Space Models
Authors:
Hermann Klein,
Max Heinz Herkersdorf,
Oliver Nelles
Abstract:
The state space dynamics representation is the most general approach for nonlinear systems and often chosen for system identification. During training, the state trajectory can deform significantly leading to poor data coverage of the state space. This can cause significant issues for space-oriented training algorithms which e.g. rely on grid structures, tree partitioning, or similar. Besides hind…
▽ More
The state space dynamics representation is the most general approach for nonlinear systems and often chosen for system identification. During training, the state trajectory can deform significantly leading to poor data coverage of the state space. This can cause significant issues for space-oriented training algorithms which e.g. rely on grid structures, tree partitioning, or similar. Besides hindering training, significant state trajectory deformations also deteriorate interpretability and robustness properties. This paper proposes a new type of space-filling regularization that ensures a favorable data distribution in state space via introducing a data-distribution-based penalty. This method is demonstrated in local model network architectures where good interpretability is a major concern. The proposed approach integrates ideas from modeling and design of experiments for state space structures. This is why we present two regularization techniques for the data point distributions of the state trajectories for local affine state space models. Beyond that, we demonstrate the results on a widely known system identification benchmark.
△ Less
Submitted 10 July, 2025;
originally announced July 2025.
-
Online and Offline Space-Filling Input Design for Nonlinear System Identification: A Receding Horizon Control-Based Approach
Authors:
Max Herkersdorf,
Oliver Nelles
Abstract:
The effectiveness of data-driven techniques heavily depends on the input signal used to generate the estimation data. However, a significant research gap exists in the field of input design for nonlinear dynamic system identification. In particular, existing methods largely overlook the minimization of the generalization error, i.e., model inaccuracies in regions not covered by the estimation data…
▽ More
The effectiveness of data-driven techniques heavily depends on the input signal used to generate the estimation data. However, a significant research gap exists in the field of input design for nonlinear dynamic system identification. In particular, existing methods largely overlook the minimization of the generalization error, i.e., model inaccuracies in regions not covered by the estimation dataset. This work addresses this gap by proposing an input design method that embeds a novel optimality criterion within a receding horizon control (RHC)-based optimization framework. The distance-based optimality criterion induces a space-filling design within a user-defined region of interest in a surrogate model's input space, requiring only minimal prior knowledge. Additionally, the method is applicable both online, where model parameters are continuously updated based on process observations, and offline, where a fixed model is employed. The space-filling performance of the proposed strategy is evaluated on an artificial example and compared to state-of-the-art methods, demonstrating superior efficiency in exploring process operating spaces.
△ Less
Submitted 3 April, 2025;
originally announced April 2025.
-
Optimized Excitation Signal Design Employing Receding Horizon Control
Authors:
Max Heinz Herkersdorf,
Oliver Nelles
Abstract:
A novel excitation signal design strategy based on a receding horizon control inspired optimization is presented. The proposed method is shown to effectively generate space-filling designs within the input space of a nonlinear dynamic process, thereby enabling sophisticated acquisition of information in previously unexplored operational areas. Additionally, the strategy can intensify the exploitat…
▽ More
A novel excitation signal design strategy based on a receding horizon control inspired optimization is presented. The proposed method is shown to effectively generate space-filling designs within the input space of a nonlinear dynamic process, thereby enabling sophisticated acquisition of information in previously unexplored operational areas. Additionally, the strategy can intensify the exploitation of specific operational areas during information gathering, offering flexibility in meeting application-specific requirements.
△ Less
Submitted 1 October, 2024;
originally announced October 2024.
-
Optimized Excitation Signal Tailored to Pertinent Dynamic Process Characteristics
Authors:
Max Heinz Herkersdorf,
Tarek Koesters,
Oliver Nelles
Abstract:
The effectiveness of data-driven techniques significantly relies on the input signal used to generate the training data. Nevertheless, there is a notable gap in research when it comes to designing excitation signals for identifying nonlinear dynamic systems, likely because of the challenges involved. Based on current knowledge, it is crucial for excitation signals to effectively capture the nonlin…
▽ More
The effectiveness of data-driven techniques significantly relies on the input signal used to generate the training data. Nevertheless, there is a notable gap in research when it comes to designing excitation signals for identifying nonlinear dynamic systems, likely because of the challenges involved. Based on current knowledge, it is crucial for excitation signals to effectively capture the nonlinearity across the entire operational area and to gather insights into the area-specific dynamic process characteristics. The Incremental Dynamic Space-Filling Design (IDS-FID) strategy designs excitation signals to achieve a space-filling distribution across the input space of a nonlinear approximator used in external dynamics modeling, gathering information throughout its operational area. Simultaneously, the approach enables for a heightened focus on either the systems steady-state or transient responses during information acquisition by altering the excitation signals dynamics, facilitating targeted insights into dynamic process characteristics.
△ Less
Submitted 1 October, 2024;
originally announced October 2024.
