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.

It is often required that identified models exhibit certain stability and dissipativity properties, e.g., passivity or ℓ2-gain. The aim of this article is to develop an unconstrained parametrization of linear parameter-varying (LPV) input–output (IO) discrete-time (DT) models that guarantees stability/dissipativity by construction, i.e., the model is stable/dissipative for any choice of model parameters. To achieve this, it is shown that any quadratically stable/dissipative DT-LPV-IO model can be generated by a mapping of transformed coefficient functions that are constrained to the unit ball. The unit ball is reparameterized through a Cayley transformation, resulting in a fully unconstrained parameterization. These results immediately apply to linear time-varying IO models. In the linear time-invariant case, an unconstrained parameterization of all stable/dissipative DT transfer functions is obtained. The unconstrained parametrization enables, among others, the use of neural network coefficient functions in LPV system identification while guaranteeing stability and dissipativity.

Hard-to-model, often nonlinear dynamics limit the tracking performance of physical-model-based feedforward control in medical interventional X-ray (IX) systems. In this article, these unknown dynamics are compensated using a physics-guided neural network (PGNN) feedforward controller. The PGNN consists of a parallel combination of a physical model describing the equations of motions of the IX and a feedforward neural network. To ensure that the neural network compensates only for the dynamics not included in the physical model, the neural network output in the subspace spanned by the physical model is penalized through regularization. As a result, the physical model serves as a baseline for performance, and the neural network contribution can be monitored or disabled. The PGNN feedforward controller is validated on an experimental IX setup, illustrating its superior performance over a physical-model-based feedforward controller.

Iterative learning control (ILC) techniques are capable of improving the tracking performance of control systems that repeatedly perform similar tasks by utilizing data from past iterations. The aim of this paper is to design a systematic approach for learning parameterized feedforward signals with limited complexity. The developed method involves an iterative learning control in conjunction with a data-driven sparse subset selection procedure for basis function selection. The ILC algorithm that employs sparse optimization is able to automatically select relevant basis functions and is validated on an industrial flatbed printer.

Ensuring stability of discrete-time (DT) linear parameter-varying (LPV) input-output (IO) models estimated via system identification methods is a challenging problem as known stability constraints can only be numerically verified, e.g., through solving Linear Matrix Inequalities. In this paper, an unconstrained DT-LPV-IO parameterization is developed which gives a stable model for any choice of model parameters. To achieve this, it is shown that all quadratically stable DT-LPV-IO models can be generated by a mapping of transformed coefficient functions that are constrained to the unit ball, i.e., a small-gain condition. The unit ball is then reparameterized through a Cayley transformation, resulting in an unconstrained parameterization of all quadratically stable DT-LPV-IO models. As a special case, an unconstrained parameterization of all stable DT linear time-invariant transfer functions is obtained. Identification using the stable DT-LPV-IO model with neural network coefficient functions is demonstrated on a simulation example of a parameter-varying mass-damper-spring system.

The performance of feedforward control depends strongly on its ability to compensate for reproducible disturbances. The aim of this paper is to develop a systematic framework for artificial neural networks (ANN) for feedforward control. The method involves three aspects: a new criterion that emphasizes the closed-loop control objective, inclusion of preview to deal with delays and non-minimum phase dynamics, and enabling the use of an iterative learning algorithm to generate training data in view of addressing generalization errors. The approach is illustrated through simulations and experiments on an industrial flatbed printer.

Cross-coupled iterative learning control (ILC) can improve the contour tracking performance of manufacturing systems significantly. This paper aims to develop a framework for norm-optimal cross-coupled ILC that enables intuitive tuning of time- and iteration-varying weights of the exact contour error and its tangential counterpart. This leads to an iteration-varying ILC algorithm for which convergence conditions are developed. In addition, a resource-efficient implementation is developed that reduces the computational load significantly and enables the use of long reference signals. The approach is experimentally validated on an industrial flatbed printer.

Identifying structured discrete-time linear time/parameter-varying (LPV) input-output (IO) models with global stability guarantees is a challenging problem since stability for such models is only implicitly defined through the solution of matrix inequalities (MI) in terms of the model's coefficient functions. In this letter, a structured linear IO model class is developed that results in a quadratically stable model for any choice of coefficient functions, enabling identification using standard optimization routines while guaranteeing stability. This is achieved through transforming the MI-based stability constraints in a necessary and sufficient manner, such that for any choice of transformed coefficient functions the MIs are satisfied. The developed stable LPV-IO model is employed in simulation to estimate the parameter-varying damping of mass-damper-spring system with stability guarantees, while a standard LPV-IO model results in an unstable estimate.

Tracking performance of physical-model-based feedforward control for interventional X-ray systems is limited by hard-to-model parasitic nonlinear dynamics, such as cable forces and nonlinear friction. In this paper, these nonlinear dynamics are compensated using a physics-guided neural network (PGNN), consisting of a physical model, embedding prior knowledge of the dynamics, in parallel with a neural network to learn hard-to-model dynamics. To ensure that the neural network learns only unmodelled effects, the neural network output in the subspace spanned by the physical model is regularized via an orthogonal projection-based approach, resulting in complementary physical model and neural network contributions. The PGNN feedforward controller reduces the tracking error of an interventional X-ray system by a factor of 5 compared to an optimally tuned physical model, successfully compensating the unmodeled parasitic dynamics.

The performance of a feedforward controller is primarily determined by the extent to which it can capture the relevant dynamics of a system. The aim of this paper is to develop an input-output linear parameter-varying (LPV) feedforward parameterization and a corresponding data-driven estimation method in which the dependency of the coefficients on the scheduling signal are learned by a neural network. The use of a neural network enables the parameterization to compensate a wide class of constant relative degree LPV systems. Efficient optimization of the neural-network-based controller is achieved through a Levenberg-Marquardt approach with analytic gradients and a pseudolinear approach generalizing Sanathanan-Koerner to the LPV case. The performance of the developed feedforward learning method is validated in a simulation study of an LPV system showing excellent performance.

Repetitive control can lead to high performance by attenuating periodic disturbances completely, yet it may amplify non-periodic disturbances. The aim of this paper is to achieve both fast learning and low errors in repetitive control. To this end, a nonlinear learning filter is introduced that distinguishes between periodic and non-periodic errors based on their typical amplitude characteristics to adapt the extent to which they are included in the repetitive controller. Convergence conditions for the nonlinear repetitive control system are derived by casting the resulting closed-loop as a discrete-time convergent system. Simulation results of the proposed approach demonstrate fast learning and small errors.

Cross-coupled iterative learning control (ILC) can achieve high performance for manufacturing applications in which tracking a contour is essential for the quality of a product. The aim of this paper is to develop a framework for norm-optimal cross-coupled ILC that enables the use of exact contour errors that are calculated offline, and iteration-and time-varying weights. Conditions for the monotonic convergence of this iteration-varying ILC algorithm are developed. In addition, a resource-efficient implementation is proposed in which the ILC update law is reframed as a linear quadratic tracking problem, reducing the computational load significantly. The approach is illustrated on a simulation example.

Unknown nonlinear dynamics can limit the performance of model-based feedforward control. The aim of this paper is to develop a feedforward control framework for systems with unknown, typically nonlinear, dynamics. To address the unknown dynamics, a physics-based feedforward model is complemented by a neural network. The neural network output in the subspace of the model is penalized through orthogonal projection. This results in uniquely identifiable model coefficients, enabling increased performance and similar task flexibility with respect to the model-based controller. The feedforward framework is validated on a representative system with performance limiting nonlinear friction characteristics.

An increasing trend in the use of neural networks in control systems is being observed. The aim of this paper is to reveal that the straightforward application of learning neural network feedforward controllers with closed-loop data may introduce parameter inconsistency that degrades control performance, and to provide a solution. The proposed method employs instrumental variables to ensure consistent parameter estimates. A nonlinear system example reveals that the developed instrumental variable neural network (IVNN) approach asymptotically recovers the optimal solution, while pre-existing approaches are shown to lead to inconsistent estimates.

Unknown nonlinear dynamics often limit the tracking performance of feedforward control. The aim of this paper is to develop a feedforward control framework that can compensate these unknown nonlinear dynamics using universal function approximators. The feedforward controller is parametrized as a parallel combination of a physics-based model and a neural network, where both share the same linear autoregressive (AR) dynamics. This parametrization allows for efficient output-error optimization through Sanathanan-Koerner (SK) iterations. Within each SK-iteration, the output of the neural network is penalized in the subspace of the physicsbased model through orthogonal projection-based regularization, such that the neural network captures only the unmodelled dynamics, resulting in interpretable models.

The performance increase up to the sensor resolution in repetitive control (RC) invalidates the standard assumption in RC that data is available at equidistant time instances, e.g., in systems with package loss or when exploiting timestamped data from optical encoders. The aim of this paper is to develop an intermittent sampling RC framework for non-equidistant measurements. Sufficient stability conditions are derived that can be verified using non-parametric frequency response function data. This results in a frequency domain design procedure to explicitly address uncertainty. The RC framework is validated on an industrial printbelt setup for which exact non-equidistant measurement data is available.