Inversion-free feedforward and feedback control of MSM based actuator with large non-smooth input hysteresis Authors accepted manuscript
Michael Ruderman, Gianluca Giostra, Matteo Sette · 2025
Read Inversion-free feedforward and feedback control of MSM based actuator with large non-smooth input hysteresis Authors accepted manuscript by Michael Ruderman, Gianluca Giostra, Matteo Sette (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-18444.txt.
Introduction
Michael Ruderman, Gianluca Giostra, and Matteo Sette design, in a 2025 control paper, position control for an actuator made of magnetic shape memory material. This alloy jumps between stable shapes, creating a huge, jagged hysteresis: the link between input and motion can vanish or explode mid-move, defeating standard feedback design.
Their feedforward compensator needs no inversion of the hysteresis model, which they build from a classical operator model. A robust proportional-integral feedback loop is then shaped around it, allowing for sensor noise and filter lag. Experiments on the real setup show the two parts working in parallel beat either alone at holding position.
Catalog record
| CREATOR | Michael Ruderman, Gianluca Giostra, Matteo Sette |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-18444
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Inversion-free feedforward and feedback control of MSM based actuator with large non-smooth input hysteresis Authors accepted manuscript Michael Ruderman, Gianluca Giostra, Matteo Sette Abstract Dynamic systems with a large and non-smooth hysteresis in the feedforward channel challenge the design of feedback control since the instantaneous input gain is varying during the operation, in the worst case between zero and infinity. Magnetic shape memory (MSM) actuators with multi-stable transitions represent such untypical system plant with only the output displacement being measured. This paper provides a case study of designing the feedforward and feedback control system for an MSM-based actuator setup with a fairly high level of the output sensing noise. First, the recently introduced inversion-free feedforward hysteresis compensator is adapted for the Krasnoselskii-Pokrovskii operator model. Then, a robust feedback proportional-integral (PI) loop shaping is performed, while taking into account the lagging behavior of the low-pass filtering and system uncertainties. Experimental results show that the parallel action of feedforward and feedback parts improves the overall performance of position control. keywords Hysteresis compensation \sep magnetic shape memory \sep 2DOF control \sep Krasnoselskii-Pokrovskii model \sep hysteresis operator \sep inversion-free feed-forwarding , , 1 Introduction Two-degree-of-freedom (2DOF) control with feedforward may be required in various actuated systems, especially where the modeled but not measurable nonlinearities cannot be compensated effectively by a feedback action only. An example of such single-input-single-output system is a magnetic shape memory (MSM) actuator, which is based on active materials with magnetic field-induced elongation and mechanical stress-driven retraction, see Ullakko 1996 . MSM actuators evoked large attention in research communities since some two decades, leading to a variety of the proposed approaches to dynamics modeling, see e.g. Sarawate and Dapino 2008 , and control, see e.g. Riccardi et al. 2013 ; Ruderman and Bertram 2013 . Over the decades, enthusiasm for innovation surrounding the MSM actuator technologies subsided due to the findings regarding their reversibility, energy efficiency, and uncertainties in the dynamic and static response. One of the biggest challenges identified in the conducted research, is a strongly pronounced nonlinear and multi-stable behavior of the MSM drive elements, that makes their accurate control a non-trivial task. Nevertheless, the latest detailed modeling studies, e.g. Ehle et al. 2023 , show a continued interest in the MSM-based actuation technologies, also with regard to different (alternative) MSM actuator topologies and structures, like for example multi-stable push-pull linear actuator, e.g. Courant and Maas 2024 . One of the most pronounced characteristic input-output properties of MSM actuators, is a large and less uniform counterclockwise input hysteresis between the exciting magnetic field and the output strain (elongation), while the latter is driven back by an applied compressive mechanical stress, cf. Sarawate and Dapino 2008 ; Ruderman and Bertram 2011 . The hysteresis , see e.g. Mayergoyz 2003 for basics, is understood as a nonlinear multi-valued but rate-independent mapping that features a nonlocal memory. Various nonlinear operator-based hysteresis models are known (see e.g. Visintin 1994 for details) that describe nonlinear phenomena with memory-affected transitions, while the so-called Preisach type hysteresis model can be seen as a certain over-class for different operator-based approaches. For a recent review on differential equations with hysteresis operators, existence of their solutions, and stability we also refer to Leonov et al. 2017 . To control the hysteresis in systems through an inverse compensation, various approaches (some of which are approximative and ad-hoc) were proposed over the decades. An overview which includes iterative hysteresis inversion for classes of Preisach, Prandtl-Ishlinskii, and Krasnoselskii-Pokrovskii operator models can be found e.g. in Iyer and Tan 2009 . Since the Preisach-type hysteresis models constitute a continuum (or a large number in a discretized setting) of the individual elementary hysteresis operators, the question of a practical implementation has always played one of the key roles for model-based hysteresis control. Examples of an efficient real-time compatible implementation of the Preisach hysteresis model are available, e.g. Ruderman 2015 . Also the robust online recursive estimation methods ( Ruderman and Rachinskii 2018 ) were proposed for identifying the otherwise unknown distribution of the elementary hysteresis operators, building up the overall input-output hysteresis map. In this paper, we address the problem of an output displacement control of the MSM-based actuator ( Ruderman and Bertram 2011 ) by developing a 2DOF control structure whose model-based feed-forwarding is based on the inversion-free hysteresis compensation proposed recently in Ruderman 2023 . The targeted MSM actuator system features a fairly high level of sensing noise (in addition to model uncertainties) that challenges significantly a feedback control part and makes an hysteresis compensating feed-forwarding indispensable. The contribution of the paper is twofold. First, we adapt and prove the inversion-free feedforward hysteresis compensation from Ruderman 2023 to the Krasnoselskii-Pokrovskii hysteresis model with non-smooth transitions and loss of strict monotonicity. Second, we demonstrate a practical experimental case study which highlights several hurdles of the essential model identification, design of the output feedback control part, and realization of the 2DOF control architecture. 2 Inversion-free feedforward hysteresis compensation 2.1 Internal model based approximation of inverting Following the internal model principle (IMP), see Francis and Wonham 1976 , that requires incorporating in the feedback path a suitably reduplicated model of the dynamic structure of disturbance, an approximation of inverting the hysteresis map H was proposed in Ruderman 2023 as it is shown in Fig. 1 . Figure 1: Inversion-free feedforward hysteresis control. Recall that attractive properties of IMP carry over also to general nonlinear case, see Economou et al. 1986 . In Ruderman 2023 , it was shown that since a hysteresis map H does not have any internal state dynamics and, thus, can be seen at each input-output operation point as a linear combination of the known bounded bias Y_{0} and finite gain \gamma=\partial y/\partial u , the \hat{y}(t)\rightarrow y^{\ast}(t) and thus u(t) approximates an inverse of the hysteresis map H:\,u\mapsto y . It was demonstrated in Ruderman 2023 that the integrator gain K can be assigned arbitrary large, theoretically, and the loop error magnitude |y^{\ast}-\hat{y}| goes to 0 at steady-state, and becomes 1 at higher frequencies of y^{\ast}(t) , provided \gamma>0 everywhere. 2.2 Krasnoselskii-Pokrovskii hysteresis model The so called Krasnoselskii-Pokrovskii (KP) hysteresis operator, see Krasnosel’skii and Pokrovskii 1989 for basics, is a memory-affected Preisach-type operator (cf. Mayergoyz 2003 ) whose input-output transitions are as depicted in Fig. 2 . Figure 2: Krasnoselskii-Pokrovskii hysteresis operator. The nonlinear static map is given by \bar{y}=h[u,y_{0},\alpha,\beta], (1) where the initial state is captured by y_{0} and the spatial parameters are the output zero-crossing values \alpha>\beta . Note that if \alpha=\beta , the the hysteresis operator ( 1 ) reduces to the saturated linear unity slope. Also we note that this zero-crossing-type parametrization only follows some conventions of the hysteresis literature (cf. e.g. Visintin 1994 , Mayergoyz 2003 ), while \alpha,\beta parameters can also be assigned for instance to the threshold values where the slope transitions (see Fig. 2 ) begin, cf. e.g. Iyer and Tan 2009 . The overall KP hysteresis model is obtained by a weighted continuum of the elementary operators ( 1 ), in accord with the Preisach-type hysteresis model principles ( Visintin 1994 ; Mayergoyz 2003 ), as y(t)=\iint\limits_{\alpha\geq\beta}\rho(\beta,\alpha)h\bigl[u(t),y_{0},\alpha,\beta\bigr]d\beta d\alpha. (2) Here \rho:P\rightarrow\mathbb{R}_{+} is a positive, integrable function on P=\bigl\{(\beta,\alpha):\>\alpha\geq\beta\bigr\} which constitutes the density