Artificial Intelligence Control Methodologies for Shape Memory Alloy Actuators: A Systematic Review and Performance...

Shape Memory Alloy (SMA) actuators are pivotal in modern engineering due to their unique thermomechanical properties, but their inherent non-linearities, hysteresis, and temperature sensitivity pose significant control challenges. This systematic review evaluates artificial intelligence (AI)-based control methodologies to address these limitations, analyzing their efficacy in enhancing precision, adaptability, and reliability for SMA and Magnetic SMA (MSMA) systems. A PRISMA-guided literature review (2003-2025) identified 24 studies, which were categorized by control architectures (hybrid AI-linear, pure AI, adaptive, and model predictive control) and evaluated through quantitative metrics, including Root Mean Square Error (RMSE%) and a weighted scoring system for experimental rigor. Results revealed hybrid AI-linear controllers as the dominant approach (36%), with online-trained neural networks achieving superior accuracy (+2.4%) over offline methods. Feedforward neural networks outperformed recurrent architectures (+3.1%), while Model Predictive Control (MPC) excelled for SMA actuators (+5.8% accuracy) but underperformed for MSMAs (-7.7%). Sensorless strategies proved advantageous for MSMAs (+5.0%), leveraging intrinsic material properties like electrical resistance for state estimation. The analysis underscores AI's capacity to mitigate hysteresis and non-linear dynamics, though material-specific optimization is critical: SMA systems favor dynamic control and MPC, whereas MSMAs benefit from sensorless AI and pure neural networks. Challenges persist in computational demands for online training and reinforcement learning's exploration-exploitation trade-offs. Future research should prioritize adaptive algorithms for fatigue compensation, lightweight AI models for embedded deployment, and standardized benchmarking to bridge material-specific performance gaps. This synthesis establishes AI as a transformative paradigm for SMA actuation, enabling precise control in aerospace, biomedical, and soft robotics applications.