Research On Disturbance Model Identification And Optimal Control Technology For Adaptive Optics Systems

Adaptive optics(AO)is a technology that obtains high-resolution images which almost close to diffraction-limited quality by overcoming the impact of dynamic wavefront distortion.So far,AO has been widely used in astronomical observation,laser transmission,high-resolution retinal imaging and so on.When we use the ground-based astronomical telescopes to perform scientific observations,the wavefront distortion caused by randomly changing atmospheric turbulence is the main reason for the degradation of the imaging quality of the system.In addition,changes of the working environment and the system platform itself will also lead to non-turbulent aberration disturbances,mainly including vibration caused by wind or other excitation sources(coolers,shutters,etc.),which greatly limits the improvement of system performance.At present,most of AO systems utlize the classic proportional-integral controllers,which cannot effectively suppress these non-turbulent disturbances,and ignore the characteristics of these disturbances which varying with the evolution of time.It is incompetent for these controllers to meet the needs of high-resolution imaging of large astronomical telescopes.To address this problem,we carry out the following researches and obtain a series of achievements about the correction of the disturbances(including turbulence and non-turbulent disturbances).1.We establish a dynamic disturbance model with practical physical meaning based on a detailed explanation of the disturbance.Firstly,an independent disturbance component model that can characterize the actual physical process has been obtained by analyzing the relevant characteristics of different perturbation compenents,namely a second-order Auto-Regressive(AR2)model.Then,we established the time domain dynamic model and the frequency domain dynamic model of the composite disturbance in the AO system according to the relationship between each perturbation compenent.The source of each perturbation component is different and independent of each other.Consequently,according to the construction of disturbance model that evoluted along with time,we provide an accurate description of the dynamic disturbance signal to prepare for the design of the disturbance controller in the AO system.2.We propose an automatic perturbation identification method which involves the spectral separation procedure and a parameter optimization process based on the particle swarm optimization(PSO)algorithm,which can be run online periodically without manually setting of any initial values of the model parameters and any auxiliary system.First of all,we segment the power spectral density curve for extracting each independent perturbation component from the multiple-frequency disturbance signal,and obtain the number of effective independent perturbation compenents during the current operating period.Secondly,we utilize the advantage of the PSO algorithm,which has random initial values,few parameters to adjustand and fast convergence speed,to effectively identify the frequency domain model parameters of the disturbance.In addition,we compared the performance of the identified results with other identification methods.The identification method has been verified on simulated synthetic disturbance,which indicates that the calculation time of the identification method proposed in this paper is greatly reduced and the identification accuracy of the proposed method is also better,due to the adjustable inertia weight to decrease the detrimental possibility of falling into local minima.It provides important support for the design of the optimal controller.3.We propose a multi-mode disturbance controller based on the linear quadratic Gaussian(LQG)algorithm with adjustable control structure to address the problem that non-turbulent disturbances in high-order modes in AO systems are difficult to correct.First,according to the parameters of the disturbance model obtained by identification method and the number of independent perturbation component,the LQG controller with adjustable control structure is designed with the minimum residual error of disturbances correction.Then,according to the number of independent perturbation compentents and the corresponding model parameters obtained by identifying each of the multi-order disturbance mode,we can obtain the control matrix dimension of the multi-mode disturbance controller and calculate the corresponding control parameter matrix to perform the LQG controller with adjustable control structure for multiple modes in adaptive optics system.Finally,we compare the performance of the LQG control on multiple modes with the optimized model gain integral(OMGI)controller.The simulation results show that the multi-mode disturbance controller proposed in this paper has good mitigation performance for the disturbance in the high-order mode.At the same time,compared with the OMGI controller,the proposed control method shows a better performance for the narrow-band disturbance in the high-order mode.Moreover,it also validated the effectiveness of the disturbance model and the reliability of the identification method.4.The identification method as well as the proposed control technology have been verified on the consecutive experimental data recorded by the 1-m New Vacuum Solar Telescope(NVST).Furthermore,we use the consecutive experimental data sets to replay the process of dynamic measurement along with time to assess the real-time feasibility of the online identification method for 1-m NVST.Then,the multi-mode disturbance controller based on the LQG algorithm has been accomplished in laboratory based on the experiment platform of AO system.Finally,we perform an experimental verification about the suppression of narrow-band perturbations caused by the coolers equiped in the 1.8m Chinese Large Solar Telescope(CLST)system platform.5.Considering the computational burden of the actual systems,we we propose a hybrid control method for the high-order correction loop,which is feasible from an implementation viewpoint.The performances of the proposed control method are evaluated with the on-sky measurement data recorded by the 1-m NVST.The results show that most of disturbances are effectively filtered by the hybrid control method,and the execution time of also obviously reduced,which is beneficial for the real-time mitigation of disturbance.Above all,based on the research on the disturbance control,we obtain the time-domain dynamic model and the frequency-domain dynamic model of compound disturbance in the AO systems,and provid an accurate description of the dynamic disturbance by the automatic perturbation identification method which involves the spectral separation procedure and a parameter optimization process based on the PSO algorithm,which address the problem that it is difficult for the current AO systems to obtain the disturbance model accurately.Besides,according to the disturbance model parameters and the number of pure-frequency perturbation components corresponding to the different modes identified,the multi-mode LQG disturbance controller with adjustable control structure is designed with the minimum residual disturbance in the AO systems,which solve the problem that disturbances,especially narrow-band disturbances,are difficult to correct in the high-order mode.It greatly improves the control performance of the AO system for disturbances mitigation.In addition,we also propose a hybrid control strategy which is suitable for the correction of high-order disturbances in actual AO systems.It reduces the calculation pressure faced by the AO system when performing high-order disturbance correction to a certain extent,and provides a solution for the application of LQG control in the actual system.