In this paper, a neural observer-based quantized output feedback control with guaranteed transient performance is developed for MEMS gyroscopes. Firstly, to generate piece-wise quantized control signals that can be performed in the digital control system, a hysteresis quantizer (HQ) is employed, meanwhile the undesirable chattering phenomena occurring universally in the traditional non-hysteresis quantizers can be also discarded. Subsequently, to provide prescribed specifications on the transient and steady-state behaviors of output tracking errors, asymmetric preselected boundaries and error transformation functions are constructed to convert the original constrained dynamics into an unconstrained one, such that the predetermined transient and steady-state performance can be achieved. Furthermore, with the aid of presented minimal learning parameter-based neural observer (MLP-NO), not only the unknown disturbances can be online identified, but also the notorious difficulties, known as the excessive occupation of the limited computational resource as well as the restriction of immeasurable velocity states, can be simultaneously circumvented. Also, the stability of resulting control law is analyzed via Lyapunov function and non-smooth analysis technique. The simulation results and comparisons validate the effectiveness of the proposed scheme.

本文针对MEMS陀螺仪开发了一种基于神经观测器的，具有保证的瞬态性能的量化输出反馈控制。首先，为了产生可以在数字控制系统中执行的分段量化控制信号，采用了磁滞量化器（HQ），同时还可以丢弃在传统的非磁滞量化器中普遍出现的不希望的颤动现象。随后，为了提供有关输出跟踪误差的瞬态和稳态行为的规定规范，构造了不对称的预选边界和误差转换函数，以将原始受约束的动力学转换为不受约束的动力学，从而可以实现预定的瞬态和稳态性能。取得成就。此外，借助提出的基于最小学习参数的神经观测器（MLP-NO），不仅可以在线识别未知干扰，而且还可以识别臭名昭著的难题，即过度占用有限的计算资源以及对算法的限制。不可估量的速度状态，可以同时被规避。此外，通过Lyapunov函数和非平滑分析技术分析了所得控制律的稳定性。仿真结果和比较验证了所提方案的有效性。通过Lyapunov函数和非光滑分析技术对所得控制律的稳定性进行了分析。仿真结果和比较验证了所提方案的有效性。通过Lyapunov函数和非光滑分析技术对所得控制律的稳定性进行了分析。仿真结果和比较验证了所提方案的有效性。