This paper deals with the stability analysis of PI and PID control of dynamic systems with an input hysteresis described by a modified Prandtl-Ishlinskii model. The problem of the asymptotic tracking of constant references is reformulated as the stability of a polytopic linear differential inclusion. This offers a simple linear matrix inequality condition that, when satisfied with the chosen PI or PID controller gains, ensures the tracking of constant references, allows the designer to establish a performance index and allows using powerful analysis and design tools for the controller. The validation of the approach is performed experimentally on a Magnetic Shape Memory Alloy micrometric positioning system.
I. INTRODUCTIONHE use of smart materials such as piezoelectric ceramics as well as magnetostrictive, thermal shape memory and magnetic shape memory alloys (MSMA) in innovative devices has generated the fairly active research area of unconventional actuators [1]. Especially in positioning applications, the phenomenon of hysteresis, which arises in almost every unconventional actuator, is known to produce poor tracking performance or even instability if not properly handled by the controller. Moreover, the rate-independent hysteresis is often coupled with a further dynamic component (for example, the mechanical load that a positioning actuator has to move), leading to a rate-dependent input-output behavior. Literature offers a wide range of solutions to control hysteretic systems. A well-known strategy is to achieve a linearization of the input-output characteristic by inverting a model of the hysteresis, and relies on the maturity of several phenomenological models [2],[3],[4]. This method can be enforced with an outer control loop to improve the performance. Moreover, the inverse model can be made adaptive if the hysteresis effects are time-varying (for instance, due to temperature effects [5]) and/or the dynamic part of the plant is unknown [6],[7]. Unfortunately, the presence of the model and/or the inverse model can lead to complex control algorithms. Other approaches do not explicitly develop a model for the hysteresis, but rather attempt to compensate its effects by proper tuning of the Manuscript received March 3, 2012.