Abstract In this paper, a fuzzy wavelet neural control scheme with improved prescribed performance is investigated for micro-electro-mechanical system (MEMS) gyroscope in the presence of uncertainties and input quantization. A hysteresis quantizer (HQ) is introduced in the controller design to generate input signal in a finite set, which can greatly reduce the actuator bandwidth without decreasing the control accuracy, and avoid the undesirable chattering occurring universally in other quantizers. To guarantee the output tracking with better prescribed transient behavior, a modified prescribed performance control (MPPC) consisting of asymmetric performance boundaries and an error transformation function is explored, such that arbitrarily small overshoot can be assured without retuning design parameters. Unlike the traditional neural network that suffers from explosion of learning, a fuzzy wavelet neural network (FWNN) based on minimal-learning-parameter (MLP) is designed to identify uncertainties with slight computational burden. A robust quantized control scheme is synthesized to compensate for quantization error and achieve prescribed ultimately uniformly bounded (UUB) tracking. Finally, extensive simulations are presented to verify the effectiveness of proposed control scheme.

中文翻译：

---

受输入量化影响的 MEMS 陀螺仪具有改进规定性能的模糊小波神经控制

摘要 本文针对存在不确定性和输入量化的微机电系统（MEMS）陀螺仪，研究了一种具有改进规定性能的模糊小波神经控制方案。在控制器设计中引入了迟滞量化器（HQ）来产生有限集合的输入信号，可以在不降低控制精度的情况下大大降低执行器带宽，并避免其他量化器普遍存在的不良抖动。为了保证具有更好规定瞬态行为的输出跟踪，探索了由非对称性能边界和误差转换函数组成的改进规定性能控制 (MPPC)，这样可以在不重新调整设计参数的情况下确保任意小的过冲。与遭受学习爆炸的传统神经网络不同，基于最小学习参数 (MLP) 的模糊小波神经网络 (FWNN) 旨在识别具有轻微计算负担的不确定性。综合了稳健的量化控制方案以补偿量化误差并实现规定的最终均匀有界 (UUB) 跟踪。最后，大量的模拟被提出来验证所提出的控制方案的有效性。