کاربرد مستمر کنترل یادگیری تکراری سوئیچینگ بدون تنظیم مجدد در یک سیستم اسکن نوری

Modern optical scanning systems often use the advanced servo system to enhance the scanning accuracy and to increase the field of view. This paper presents a continuous no-reset switching iterative learning control algorithm for a novel optical scanning system to achieve the requirements of both fast response and wide field of view. In addition, the proposed method overcomes the non-converging reset error problem experienced by most conventional iterative learning control algorithms.

Modern optical scanning systems emphasize on the scanning range, scanning speed and the scanning resolution. A smart mechanism with very high servo performance is most suitable for these uses; therefore, it has become very popular in the optical scanning applications. Lately, a new step-and-scan path in Fig. 1 has been from based on the knowledge of the previous operations of the same task [2]. Therefore, this study has also chosen the ILC to achieve the high performance servo requirements.It is noted that the ILC algorithm, assumes that the initial state of the plant is equal to that of the desired trajectory so that perfect tracking can be achieved. While it is very difficult to assign the initial states in practice, many studies had discussed the design of robust ILC input against initial state errors [3]. There are different structures proposed for fixed initial state error or bounded variable initial state error [4], [5], [6] and [7], even so all the articles have to constrain the initial states to within a region. In 1996, Sison and Chong brought forth an interesting concept of no-rest iterative learning control (NRILC) [8] to convert the iterative procedure in the linear time invariant single-input–single-output (SISO) system into an equivalent system for stability analysis. The stabilizing control in their paper is not structured for practical application and they did not discuss the issue of initial state error nor offer any simulation or experiment results. This paper first extends the conventional ILC setting to the MIMO situation for the application to a novel high-speed optical scanning mechanism system. A continuous iterative learning control (CILC) algorithm is then proposed to suppress the initial state error that often diverges upon ILC. The system architecture and identification results are depicted in Section 2. The traditional ILC control criteria are briefly described in Section 3. This section also proposes the use of the traditional ILC control criterion to simulate the NRILC response, termed the continuous iterative learning control (CILC) to distinguish the NRILC in [8]. The original CILC produces poor system responses because it violates the assumptions for the conventional ILC. Section 3.2.2 then presents a modified CILC strategy noticed in several optical systems, particularly for the long distance and large field scanning systems. There are 4n steps in the step-and-scan path. This scanning path is usually arranged for an array of charge coupled devices (CCD) or complementary metal oxide semiconductor (CMOS) detectors. The signal from a single CCD or CMOS detector image displays only a small portion of the front view, and the full view is reconstructed by stitching the images together. To reconstruct a truthful front image, the stepping motion requires not only a precise position control but also a mechanism that can provide a large acceleration upon request. The conventional high torque actuators and its driving mechanisms always take up huge spaces. The proposed novel optical scanning system [1], on the other hand, uses piezo-electric transducers (PZT) with flexure joints and is very compact in size. From the servo point of view, the step-and-scan motion requires sufficient control sampling points during each step; however, there is a hardware limit on the number of usable sampling points. It is both unwise and uneconomic to change the hardware to accommodate for the growing step number n. Thus, making use of the previous control information for self control tuning would be the best way to suppress the tracking error. There are mainly two types of learning control that serves the purpose: the repetitive control (RC) and the iterative learning control (ILC). The RC design eliminates the periodic system disturbances and periodic steady state error, but it along does not improve the transient tracking performance. There are many modifications to the RC methods to enhance the steady state error performance, but they all suffer from the compromises between achievable phase compensation and system instability. The ILC is an effective method for reducing tracking errors from repeated operations. The ILC is robust against model uncertainties; it requires no detailed prior knowledge about the system; it modifies an unsatisfactory control input overcome the reset assumption. The modified CILC strategy is then extended to the MIMO case. The simulation outcomes in Section 3.2.3 verify that the modified CILC strategy is efficient. The section then discusses the non-zero initial error problem for some various optical servo systems. Section 4 displays the experimental results. Both simulation and experimental results of this application confirm that the proposed method can effectively suppress the iterative learning error and achieve the desired performance.

This paper proposed a CNRSILC algorithm for a high performance novel optical scanner. The system represented a typical ILC application and suffered the common ILC problem of diverging initial error. This paper first presented an MIMO version of the ILC control constraints, and used computer simulations to verify the analysis. The simulation also demonstrated the effects of accumulated initial error. The paper then moved on to propose the continuous no-reset switching iterative learning control algorithm. Although a rigorous stability analysis is not yet complete, the application of CNRSILC on several other servo mechanisms all proved that the method was effective in relaxing the reset assumption on the initial conditions. Both the simulation results and the experimental results confirmed the performance of the proposed control. The experiment on the step-and-scan tests further demonstrated the advantage of using the CNRSILC.