Recent Advances in Steerable Constrained Output-Feedback Control for MEMS Gyroscopes Using Steerable Graph Neural Networks with Limited Transmission Bandwidth: A Systematic Review

Abstract

Microelectromechanical systems (MEMS) gyroscopes are essential sensing devices used in aerospace, autonomous vehicles, robotics, consumer electronics, and industrial automation. However, their performance is significantly affected by fabrication imperfections, thermal drift, nonlinear dynamics, and mechanical cross-coupling, which reduce measurement accuracy under practical operating conditions. This review examines the integration of constrained output-feedback control, steerable graph neural networks (SGNNs), and bandwidth-efficient networked control for intelligent MEMS gyroscope systems. Modern output-feedback control methods, including adaptive backstepping, sliding mode control, barrier Lyapunov functions, and model predictive control, improve system stability while relying only on measurable outputs. SGNNs further enhance these controllers by representing multi-axis gyroscope structures as graphs and exploiting rotational and directional symmetries through equivariant learning, enabling accurate nonlinear state estimation and improved robustness. Practical communication constraints are addressed using event-triggered transmission, self-triggered control, quantization, and packet-loss compensation to ensure reliable operation over limited bandwidth networks. Based on more than twenty-five representative studies involving MEMS gyroscope platforms, inertial measurement units, and synthetic datasets, this review demonstrates that SGNN-enhanced constrained output-feedback controllers consistently outperform conventional adaptive and model-based approaches in tracking accuracy, disturbance rejection, robustness, and communication efficiency, highlighting their potential for next-generation intelligent inertial sensing and embedded control systems.