Model Predictive Control for an SMA Actuator Based on an SGPI Model

Wei Liu, Houzhen Wei, Yan Pang, Xudong Tang, Kai Wang, Wenya ZhouAerospace 2026 · DOI

TL;DR

This paper tackles the challenge of precisely controlling Shape Memory Alloy (SMA) actuators, which suffer from complex hysteresis and noise. They introduce a new "Smoothed Generalized Prandtl-Ishlinskii" (SGPI) model for SMA hysteresis, combine it with an Extended Kalman Filter (EKF) for state estimation, and use Model Predictive Control (MPC) for high-precision tracking. The EKF-MPC approach demonstrates superior performance in simulations compared to traditional PID and feedforward methods.

What Problem Does This Paper Address?

Shape Memory Alloy (SMA) actuators are attractive for aerospace applications due to their large deformation and simplicity. However, their inherent hysteresis—a phenomenon where the output depends not only on the current input but also on its past history—makes precise control extremely difficult. This hysteresis is often saturated and asymmetric, meaning the actuator doesn't behave the same way when heating up versus cooling down, or when moving in one direction versus another. This, combined with process and measurement noise, severely limits their use in high-precision systems. Existing models often struggle to balance accuracy, smoothness for control design, and computational efficiency.

Methods

The core of this work lies in a novel approach combining three key components:

  1. Smoothed Generalized Prandtl–Ishlinskii (SGPI) Model: The authors propose a modification to the Generalized Prandtl–Ishlinskii (GPI) hysteresis model. The standard GPI model uses non-smooth "max" and "min" operators, which are problematic for control design as they lead to non-differentiable system dynamics. The SGPI model replaces these with "smoothed" versions, introducing a small parameter ($\epsilon$) that allows for analytical differentiability while closely approximating the original hysteresis behavior. This makes the model suitable for state-space representation.
  2. Extended Kalman Filter (EKF): Since the internal hysteresis states of the SMA actuator are not directly measurable, an EKF is employed. The EKF uses the SGPI model and the measurable outputs (temperature and stiffness coefficient) to estimate the unmeasurable internal states in real-time, even in the presence of noise. This provides the necessary state information for the controller.
  3. Model Predictive Control (MPC): With the estimated states from the EKF, an MPC controller is designed. MPC uses a receding-horizon optimization approach to predict future system behavior and compute optimal control inputs that minimize a cost function while satisfying constraints. The paper uses the control increment ($\Delta u$) as the optimization variable, which inherently introduces integral action, improving robustness against model mismatches and parameter drifts.

Datasets

Not discussed in the paper. The study relies on experimental data collected from an SMA actuator setup to identify model parameters and validate the proposed EKF-MPC framework through simulations.

Key Insights and Ideas

  • Differentiable Hysteresis Modeling is Crucial for Control: The non-differentiability of traditional hysteresis models is a major bottleneck for advanced control and state estimation. The SGPI model's smooth approximation is key to enabling these techniques.
  • EKF is Essential for Unobservable States: SMA hysteresis is complex and internal states are hidden. The EKF effectively bridges this gap, providing the necessary information for the MPC.
  • MPC's Proactive Nature Outperforms Reactive Controllers: Unlike PID or feedforward methods that react to errors, MPC's predictive capability allows it to anticipate future needs and proactively adjust control inputs, especially crucial for compensating thermal inertia.
  • Control Increment Improves Robustness: Using the control increment ($\Delta u$) in MPC, rather than the absolute input, introduces integral action, making the system more resilient to uncertainties and disturbances.

Key Results

The paper presents simulation results demonstrating the effectiveness of the proposed EKF-MPC approach:

  • Improved Tracking Performance: The EKF-MPC controller achieved superior stiffness coefficient tracking compared to PID and feedforward control across multiple step transitions.
  • Reduced Overshoot: EKF-MPC showed significantly lower overshoot (within 2% in some steps) compared to PID (up to 31%) and feedforward (up to 91%).
  • Faster Settling Time: The average settling time for EKF-MPC was 6.38 seconds, substantially faster than PID (11.10 s) and feedforward (17.38 s).
  • Lower Steady-State Error: EKF-MPC achieved the lowest average steady-state error (0.1355 N/m).
  • Robustness to Noise and Initial Conditions: The EKF demonstrated effective estimation of both measurable and unmeasurable states under noisy conditions, with significant reductions in estimation error. It also showed strong robustness to initial state guesses, converging within approximately 0.7 seconds.
  • Parameter Identification: The SGPI model, identified using Particle Swarm Optimization (PSO), showed improved accuracy over the GPI model, with RMSE, AAE, and MRDE reductions of 18.9%, 15.9%, and 36.0%, respectively.

Limitations and Future Work

  • Nominal Load Validation: The current validation was performed under a nominal load. The authors suggest that future work should focus on rigorous experimental validation under varying dynamic loads.
  • Experimental Validation: While simulations are promising, real-world experimental validation is the next crucial step to confirm the EKF-MPC framework's performance in practice.
  • Parameter Tuning: While the paper demonstrates robustness to process noise parameter variations, further investigation into optimal tuning strategies for the MPC weights (Q and R) could be beneficial.

Summary updated

Abstract

Shape memory alloy (SMA) actuators possess unique advantages for aerospace applications, including significant deformation, a high work-to-weight ratio, and structural simplicity. However, SMA actuators exhibit inherently strongly saturated and asymmetric hysteresis characteristics, which cause significant hysteresis in the output response. These hysteresis nonlinearities, compounded by process and measurement noise, severely degrade control precision. To overcome these issues, this study proposes a Smoothed Generalized Prandtl–Ishlinskii (SGPI) model to characterize such hysteresis behavior. Based on the SGPI model, we developed a state-space representation for the SMA actuator. Furthermore, an Extended Kalman Filter (EKF) is employed to estimate unmeasurable internal hysteresis states, and these estimates are subsequently utilized as input states for Model Predictive Control (MPC). The simulation results demonstrate that the proposed EKF-MPC approach achieves both rapid dynamic response and high-precision tracking control, effectively compensating for hysteresis nonlinearity while rejecting noise disturbances.
BibTeX
@inproceedings{liu2026model,
  author = {Wei Liu and Houzhen Wei and Yan Pang and Xudong Tang and Kai Wang and Wenya Zhou},
  title = {{Model Predictive Control for an SMA Actuator Based on an SGPI Model}},
  booktitle = {Aerospace},
  year = {2026},
  doi = {10.3390/aerospace13020112},
}
Figure