MEMS Gyroscopes: Advancements and Strategic Challenges#

Micro-electromechanical systems (MEMS) gyroscopes have emerged as transformative components in the field of inertial sensing, largely due to their compelling advantages, including diminutive size, minimal power consumption, inherent cost-effectiveness through batch fabrication, and straightforward integration into multifaceted systems (Gill et al., 2022; Passaro et al., 2017). Research in this area is intensely focused on enhancing critical performance metrics, including improving bias stability, reducing Angle Random Walk (ARW), and augmenting robustness against demanding environmental factors (Gill et al., 2022; Trusov et al., 2011; Hu et al., 2025).

Key innovations feature the integration of sophisticated mechanical amplification structures to bolster sensitivity (Hu et al., 2025; Li et al., 2019) or the realization of multi-DOF micro-gyroscopes based on novel electromechanical principles (Zhang et al., 2022). Such advancements present both substantial opportunities for enhanced performance and new, intricate challenges for their control and system integration (Acar & Shkel, 2008).

Need for Advanced Control and Scope of This Paper#

The seamless integration and effective, high-performance control of these advanced micro-gyroscopes within dynamic 3-DOF table systems present considerable technical challenges (Slotine & Li, 1991). These include compensating for system nonlinearities, dynamic cross-coupling effects, and sensor-specific error characteristics like noise, thermal drift, and mode-mismatch in resonant gyroscopes (Gill et al., 2022; Hu et al., 2025; Acar & Shkel, 2008).

Beyond conventional PID controllers, there is a clear need for advanced control strategies such as adaptive or robust control (Xin et al., 2022; Xu et al., 2024; Slotine & Li, 1991). Rigorous validation of these novel algorithms under realistic dynamic conditions is crucial (Liu et al., 2023; Passaro et al., 2017). This paper addresses these challenges by presenting the design, implementation, and preliminary evaluation of a comprehensive control system for a 3-DOF motion table, architected as a high-fidelity testbed for advanced micro-gyroscope control algorithms (Hancer et al., 2018).

The focus is on the judicious selection of hardware, development of advanced control software for precise multi-axis motion, and an intuitive user interface (Ogata, 2010). The ultimate goal is to establish a robust, high-performance motion control infrastructure to facilitate experimental investigation into cutting-edge dynamic stabilization techniques and validate novel control methodologies for next-generation micro-gyroscopes (Xin et al., 2022; Xu et al., 2024; Liu et al., 2023).

System Description and Mathematical Modeling#

The system under study is a three-degree-of-freedom (3-DOF) motion table designed for precise emulation of rotational movements and evaluation of advanced control algorithms for dynamic stabilization and micro-gyroscope guidance. The mechanical structure comprises three orthogonal gimbals providing roll ($\phi$), pitch ($\theta$), and yaw ($\psi$) movements. Each axis is driven by a high-precision brushless DC motor equipped with a high-resolution encoder for accurate angular position feedback.

For precise measurement of angular velocities and orientation, an advanced MEMS-based Inertial Measurement Unit (IMU) is utilized, featuring a three-axis micro-gyroscope with enhanced sensitivity, for example through mechanical levering (Li et al., 2019) or optimized resonant design (Trusov et al., 2011; Hu et al., 2025), and a three-axis accelerometer. The sensor selection considered key parameters such as low bias stability, low noise, and adequate bandwidth.

The dynamic model of the 3-DOF table is derived using the Newton-Euler method (Hancer et al., 2018). Assuming gimbals are designed to minimize dynamic coupling effects, or these are compensated by the controller, the dynamics of each axis $k \in \{\phi, \theta, \psi\}$ can be approximated by:

Equation 1 - Axis Dynamics

where $J_k$ is the equivalent moment of inertia about axis $k$, $B_k$ is the equivalent viscous damping coefficient, $\alpha_k$ is the angular position of axis $k$, $\tau_{k,ctrl}$ is the control torque applied by the motor of axis $k$, and $\tau_{k,dist}$ is the sum of external and internal disturbance torques on axis $k$. Parameters $J_k$ and $B_k$ are obtained via CAD analysis and system identification experiments.

The output of the micro-gyroscope, $\omega_{gyro,k}$, is affected by various errors, modeled as:

Equation 2 - Gyroscope Error Model

Here, $\omega_{actual,k}$ is the true angular velocity of the table about axis $k$, $S_k$ is the scale factor error, $b_k$ is the gyro bias/drift, which can be a function of temperature and time, and $n_k$ is the random measurement noise, including Angle Random Walk (ARW) and Rate Random Walk (RRW) (Gill et al., 2022).

Advanced Control Strategy#

To achieve high performance in dynamic stabilization and precise trajectory tracking, an Adaptive Sliding Mode Control (ASMC) strategy is designed (Slotine & Li, 1991). The sliding surface $s_k$ for axis $k$ is defined as:

Equation 3 - Sliding Surface

where $e_k = \alpha_{k,ref} - \alpha_k$ is the angular position tracking error and $\lambda_k > 0$ is a design coefficient. The control law is chosen as:

Equation 4 - ASMC Control Law

where $\hat{J}_k$ and $\hat{B}_k$ are estimates of the system’s dynamic parameters, $\eta_k > 0$ is the gain of the robust control term, and $\text{sat}(\cdot)$ is a saturation function with boundary layer thickness $\Phi_k$ to mitigate chattering. An adaptation law is used to update $\eta_k$ or the parameter estimates online to enhance robustness against uncertainties and disturbances.

Implementation and Testbed#

The control system is digitally implemented on a real-time computer system using MATLAB/Simulink and Real-Time Workshop. Command signals are sent via DACs to motor drivers, and feedback from encoders and the IMU is acquired via ADCs and digital interfaces, such as SPI for the IMU. The inner loop for torque/velocity uses a sampling rate of 10 kHz, and the outer loop for position/stabilization uses a sampling rate of 1 kHz. Experiments include tracking standard reference trajectories, including step and sinusoidal inputs, and applying external disturbances to evaluate stabilization capabilities.

Baseline Performance with PID Controller#

The system was first tested with a classical PID controller, with gains tuned using the Ziegler-Nichols method followed by manual fine-tuning. The final gains for the roll axis are listed in Table 1.

Table 1 - Tuned Baseline PID Controller Gains for Roll Axis

Figure 2 shows the time response of the roll axis to a $20^\circ$ step input using the baseline PID controller. The system exhibits a rise time of approximately $0.8$ seconds and an overshoot of about $18\%$. The steady-state error settles to less than $2\%$ of the final value after approximately $2.5$ seconds.

Figure 2 - Step Response with Baseline PID Controller

Time response of the roll axis to a $20^\circ$ step input with the baseline PID controller. This plot displays the roll angle in degrees versus time in seconds for a reference step input. The reference is shown as a dashed line, and the actual system response as a solid line, illustrating key transient response parameters such as rise time, overshoot, and settling time.

Performance with Adaptive Sliding Mode Controller (ASMC)#

The ASMC, designed according to Equation 3 and Equation 4 with design parameters $\lambda_{\phi}=15$ and an adaptive $\eta_{\phi}$, using an initial value of $0.5$, was then implemented and tested. Figure 3 compares the roll axis time response to a sinusoidal input, $15^\circ \sin(\pi t)$, between the baseline PID controller and the ASMC.

Figure 3 - Sinusoidal Tracking Comparison: PID vs. ASMC

Comparison of roll axis time response to a $15^\circ \sin(\pi t)$ sinusoidal input with baseline PID and ASMC. This figure shows three curves against time: the sinusoidal reference trajectory, the system response with the baseline PID controller, and the system response with the proposed ASMC. This comparison evaluates tracking accuracy and error reduction for a time-varying input signal, where ASMC demonstrates significantly reduced tracking error and a smoother response.