Neural Adaptive Control for MEMS Gyroscope With Full-State Constraints and Quantized Input

In this article, we investigate the neural adaptive quantized control problem for microelectromechanical system (MEMS) gyroscope with full-state constraints and lumped disturbances. With two different kinds of one-to-one nonlinear mappings, the traditional gyroscope model with matched disturbances is transformed into an unconstrained one with both unmatched and matched disturbances, thus the predefined time-varying state constraints imposed on MEMS gyroscope can be achieved. To compensate for the lumped disturbances, a state estimator-based minimal learning parameter neural network is proposed to obtain fast and smooth disturbance estimates for both position and velocity control loops, which not only can eliminate the poor transient behaviors that widely appear in the available neural adaptive control with a large adaptive gain, but also greatly reduce the number of leaning parameters updated online. Furthermore, by employing a hysteresis logarithmic quantizer, the neglected difficulty, named as constrained data bandwidth of actuator can be overcome with less chattering in control signal, which is more convenient to implement. Finally, the neural adaptive control for MEMS gyroscope is developed such that satisfactory tracking performance is achieved despite of large disturbances, full-state constraints as well as quantized input. The effectiveness and advantages of the proposed control method are demonstrated through extensive simulations.