Abstract
Inertially stabilized platform (ISP) is an important stabilization component and has been widely used across areas because of its accurate pointing. Typically, conventional ISP systems possess accuracy of 0.35 mrad, which is not sufficient for long-distance applications. This results in the motivation of a new approach that can meet the requirements of such use cases. This dissertation focus on 1) sensing the disturbance after primary compensation of an inertially stabilized platform, where a fiber gyroscope is utilized to estimate multiple-frequency vibration motion, and 2) designing a robust and accurate tracking control algorithm for piezo-driven stages to actively attenuate the unwanted motion with an opposite but equal movement. To achieve this goal, both sensing (fiber gyroscope) and trajectory control (piezo-driven stage) system have to be designed to achieve high precision and accuracy. The main challenge for piezo-driven stages is the existence of hysteresis nonlinearity which poses difficulty for trajectory controller design. Generally, to eliminate hysteresis, Prandtl-Ishlinskii (PI) hysteresis model is used to model hysteresis and its inverse model is applied for linearizing. However, hysteresis is rate-dependent. Although PI model can be extended to model the rate-dependent behavior, its inverse model may not exist when the hysteresis curve’s gradient is not positive definite (negative gradient is likely to occur when driven at high frequency). Therefore, a neural network-based direct inverse model is proposed to tackle this problem. The benefits include 1) there is no requirement for an explicit relationship between the input rate and the weight of PI model, since RBFNN has the ability to learn any continuous and smooth function, 2) the operating bandwidth can be extended. To further improve the robustness, a proportional-integral controller with disturbance observer is integrated. The computational resource is another issue when implementing an algorithm. To achieve synchronization and high-frequency control bandwidth, the FPGA is preferred. Neural network-based approach requires more resource and is harder to be implemented for FPGA target. Thus, a simple PI hysteresis model augmented with a sliding mode control (SMC) is proposed to achieve high tracking performance while maintaining robustness. The chattering problem of SMC is inherently avoided through the uncertainty and disturbance estimator (UDE). In addition, the adaptive version is introduced to enhance its adaptability. The benefits of the proposed approach are 1) it is inherently chattering-free, 2) it does not require a reference model for UDE, and 3) it can track both periodic motion and non-periodic motion. Generally, modeling is time-consuming and costly. To reduce the modeling component, a hysteresis-free approach is proposed while maintaining the tracking accuracy, where an enhanced sliding mode control with uncertainty and disturbance estimator (SMC-UDE), featuring a PID-type sliding surface, is introduced for the motion tracking control of piezoelectric-driven stages. A disturbance observer is utilized to estimate the external disturbance and improve tracking performance. The strengths of the proposed method lie in 1) the proposed method is easily applied in the real scenario as there is no need to know the detailed information of the hysteresis (which is treated as disturbance), and 2) the proposed method is chattering-free and does not require knowing the bounds of the system uncertainties. To sense the involuntary motion, like vibration, and noise after the pan-tilt platform, a gyroscope is used. However, the gyroscope has the drift problem if numerical integration is applied directly. On the other hand, the hardware (inclusive of both sensors and actuators) and software (linear filters) utilized for involuntary motion estimation and compensation introduce unknown and time-varying phase delay, which is challenging to deal with. Hence, to address the concern of phase difference and drift, the author proposes an efficient and easy-for-deployment estimation approach to first sense the involuntary displacement motion with minimal phase difference using a gyroscope. To counteract the effect of undesired motion, a fast steering mirror is actuated to provide an equal, but an opposite movement. The fast steering mirror is driven by piezoelectric actuators and the hysteresis nonlinearity is mitigated by the proposed chattering-free sliding mode controller. Through these treatments, the proposed method has the following features: 1) minimal phase difference; 2) ability to handle multiple dominant frequencies; 3) requiring less computational load; and 4) no requirements for pre-training compared to other methods.