Many industrial motion systems require performing a variety of tasks with high precision and safety. Iterative learning control (ILC) is a method with convergent update laws, generally classified into: 1) parametrized learning approach for achieving task-flexibility against varying tasks; or 2) signal-based learning approach which can achieve perfect tracking-performance for repeating tasks. The aim of this study is to join the distinct ILC frameworks, achieving all desirable properties in a single framework. Specifications on convergence, tracking-performance and task-flexibility of the developed joint parametrized/signal-based ILC are theoretically derived, confirmed with experimental results on a two-mass system.

Factors like growing data availability and increasing system complexity have sparked interest in data-driven predictive control (DDPC) methods like Data-enabled Predictive Control (DeePC). However, closed-loop identification bias arises in the presence of noise, which reduces the effectiveness of obtained control policies. In this paper we propose Closed-loop Data-enabled Predictive Control (CL-DeePC), a framework that unifies different approaches to address this challenge. To this end, CL-DeePC incorporates instrumental variables (IVs) to synthesize and sequentially apply consistent single or multi-step-ahead predictors. Furthermore, a computationally efficient CL-DeePC implementation is developed that reveals an equivalence with Closed-loop Subspace Predictive Control (CL-SPC). Time marching simulations of DeePC and CL-DeePC are conducted using Hankel matrices of past data that are updated at every time step to induce potentially troublesome closed-loop correlations between inputs and noise. Compared to DeePC, CL-DeePC simulations demonstrate superior reference tracking, with a sensitivity study finding a 48% lower susceptibility to noise-induced reference tracking performance degradation.

Iterative learning control (ILC) is an intelligent control methodology for tackling iteration-invariant exogenous inputs. It is of great significance to develop its extrapolation for more general repetitive tasks with mutual similarity, e.g., tasks with different time scales. In practice, discrete-time ILC with sampling behavior for time-scale-varying tasks suffers from the failure of perfect corresponding learning and environment-dependent iteration-varying disturbances. This paper develops a novel direct data-based ILC algorithm using off-policy Q-learning for tasks with varying time scales, enabling the robust learning of an optimal ILC policy from experimental input/output (I/O) data. From a two-player zero-sum game perspective, the iteration-varying disturbance generated from the varying time scales of repetitive tasks is tackled quantitatively with a preset disturbance attenuation level. Further, to emphasize the importance of theoretical guarantees of reinforcement learning (RL)-based ILC designs, the data efficiency of the developed algorithm is enhanced based on Willems’ Fundamental Lemma, and a rigorous convergence analysis is given. The simulation model of an F-16 aircraft autopilot is employed to show the effectiveness of the developed approach.

Feedforward motion control for unmeasured performance variables at the point of interest is crucial for attaining high throughput and accuracy in motion systems. The aim of this paper is to develop a data-driven approach for feedforward tuning that addresses the true performance at the point of interest. The presented approach is a novel methodology that employs rational feedforward structures for performing flexible tasks with high accuracy, in conjunction with an sensor fusion for addressing the point-of-interest. In particular, the tracking error of the unmeasured performance variable is accurately estimated by combining acceleration measurements and encoder measurements. Simulation results show that optimizing for the estimated point-of-interest error achieves similar tracking performance as optimizing for the true point-of-interest error, indicating accurate sensor-fusion estimates for feedforward control. Experimental validation demonstrates that optimizing for the estimated point-of-interest error significantly reduces the estimated point-of-interest tracking error compared to minimizing the encoder error.

Physically interpretable models are essential for next-generation industrial systems, as these representations enable effective control, support design validation, and provide a foundation for monitoring strategies. The aim of this paper is to develop a system identification framework for estimating modal models of complex multivariable mechanical systems from frequency response data. To achieve this, a two-step structured identification algorithm is presented, where an additive model is first estimated using a refined instrumental variable method and subsequently projected onto a modal form. The developed identification method provides accurate, physically-relevant, minimal-order models, for both generally-damped and proportionally damped modal systems. The effectiveness of the proposed method is demonstrated through experimental validation on a prototype wafer-stage system, which features a large number of spatially distributed actuators and sensors and exhibits complex flexible dynamics.

Data-driven estimation of system norms is essential for analyzing, verifying, and designing control systems. Existing data-based methods often do not capture the inter-grid and transient behavior of the system, leading to inaccurate and unreliable system norm estimations. This paper presents a unified approach for accurate and reliable estimation of the H₂ and H_∞ norm with a limited amount of data. The key step is to exploit local parametric models that explicitly incorporate the inter-grid and transient dynamics. The system norm is estimated through the computation of local system norms of the local parametric models within their the local frequency interval. Simulation and experimental results illustrate the effectiveness of the proposed method.

The increasing demands for high accuracy in mechatronic systems necessitate the incorporation of parameter variations in feedforward control. The aim of this paper is to develop a data-driven approach for direct learning of parameter-varying feedforward control to increase tracking performance. The developed approach is based on kernel-regularized function estimation in conjunction with iterative learning to directly learn parameter-varying feedforward control from data. This approach enables high tracking performance for feedforward control of linear parameter-varying dynamics, providing flexibility to varying reference tasks. The developed framework is validated on a benchmark industrial experimental setup featuring a belt-driven carriage.

The kernel-based inverse system identification framework enables accurate identification of systems with non-minimum phase dynamics, greatly expanding the potential of non-causal system identification approaches. The existing kernel-based inverse system identification method performs the estimation assuming noisy input data, while in practice noise is typically present only in the output measurements. To address this impracticality, we propose a Bayesian identification method that employs the Expectation-Maximization algorithm and the Markov chain Monte Carlo method to enable direct identification of the inverse system using the available data. Through numerical simulations, we found that the proposed method allows for an accurate estimation of inverse models, and outperforms an indirect approach in both model fit and variance. The proposed method can be used to develop enhanced data-driven feedforward control methods that allow for flexible design while incorporating design specifications.

Iterative feedback tuning (IFT) enables the tuning of feedback controllers using only measured data to obtain the gradient of a cost criterion. The aim of this paper is to reduce the required number of experiments for MIMO IFT. It is shown that, through a randomization technique, an unbiased gradient estimate can be obtained from a single dedicated experiment, regardless of the size of the MIMO system. The gradient estimate is used in a stochastic gradient descent algorithm. The approach is experimentally validated on a mechatronic system, showing a significantly reduced number of experiments compared to standard IFT.

Multivariable parametric models are essential for optimizing the performance of high-tech systems. The main objective of this paper is to develop an identification strategy that provides accurate parametric models for complex multivariable systems. To achieve this, an additive model structure is adopted, offering advantages over traditional black-box model structures when considering physical systems. The introduced method minimizes a weighted least-squares criterion and uses a refined instrumental variable method to solve the estimation problem, achieving local optimality upon convergence. Experimental validation is conducted on a prototype wafer-stage system, featuring a large number of spatially distributed actuators and sensors and exhibiting complex flexible dynamic behavior, to demonstrate the effectiveness of the proposed method.

Mechanical ventilators are complex mechatronic devices that are essential for patients who are unable to breathe independently. The aim of this paper is to develop a systematic control method that achieves accurate tracking of both the pressure and flow to ensure comfortable breathing for the patient. This is achieved by using a feedback design procedure technique based on static decoupling and the factorized Nyquist criterion. Furthermore, switching controllers are introduced that allow for improved baseflow tracking performance. The presented control method is implemented in a real ventilator and it is demonstrated that the tracking performance is improved by conducting an experimental case-study.

Safe deployment of neural networks to classify time series in safety-critical applications relies on the ability of the classifier to detect data that does not originate from the same distribution as the training data. The aim of this paper is to propose a framework for detecting whether time series data is sampled from a different distribution than the training data, known as the problem of out-of-distribution (OOD) detection. We propose a novel distance-based OOD method for time series data using a hierarchical clustering method together with dynamic time-warping to measure the difference between a new data instance and the training set. The method is evaluated in the context of mechanical ventilation, a safety critical application, using both simulated and clinical datasets. Results of the mechanical ventilation use case demonstrate that the proposed approach effectively detects out-of-distribution data and improves classification performance in diverse settings.

Frequency Response Function (FRF) measurements are widely used in gravitational wave detectors, e.g., for the design of controllers, calibrating signals, and diagnosing problems with system dynamics. The aim of this paper is to present GraFIT: a toolbox that enables fast, inexpensive, and accurate identification of FRF measurements for gravitational wave detectors compared to the commonly used approaches, including common spectral analysis techniques. The toolbox builds upon recent developments in non-parametric system identification by utilizing a local modeling technique, which is particularly effective in reducing the impact of transient effects. It is furthermore designed to be user-friendly, handling systems of arbitrary input-output dimensions, and systems operating in a closed loop. The toolbox is validated on two experimental case studies of the Virgo detector, illustrating more than a factor 3 reduction in the estimated standard deviation using GraFIT for the same measurement times and comparable estimated standard deviations with up to ten times less data using GraFIT with respect to the commonly used spectral analysis method. As a result, GraFIT can reduce commissioning time and detector downtime due to noise injections, while also improving the overall quality of the measurements.

Online system identification algorithms are widely used for monitoring, diagnostics and control by continuously adapting to time-varying dynamics. Typically, these algorithms consider a model structure that lacks parsimony and offers limited physical interpretability. The objective of this paper is to develop a real-time parameter estimation algorithm aimed at identifying time-varying dynamics within an interpretable model structure. An additive model structure is adopted for this purpose, which offers enhanced parsimony and is shown to be particularly suitable for mechanical systems. The proposed approach integrates the recursive simplified refined instrumental variable method with block-coordinate descent to minimize an exponentially-weighted output error cost function. This novel recursive identification method delivers parametric continuous-time additive models and is applicable in both open-loop and closed-loop controlled systems. Its efficacy is shown using numerical simulations and is further validated using experimental data to detect the time-varying resonance dynamics of a flexible beam system. These results demonstrate the effectiveness of the proposed approach for online and interpretable estimation for advanced monitoring and control applications.