Research Of Stochastic Adaptive Dual Control Algorithm

With the improvement of the economic and technology in science, systems become more and more complex in such areas:aerospace systems and industrial production system, thus there are many characteristics in complex system, in which uncertainty is one of important characteristics. Traditional random theory merely considers the external noise in system, as the unknown parameter when complex system have long running time or in different environment, traditional methods can not effectively control when system in such situation. Thus for the stochastic system with unknown parameters, which was named dual uncertainty system, we need to design a new controller:on the one hand,it can make the output of system effectively track the target; On the other hand, it can identify and study the unknown parameters. so the design of the controller to make a trade-off between the two aspects of tracking and identification. Based on the parameters of different models:double uncertain systems, multiple models of stochastic system and turn-off phenomenon, the main research contents are as follows:.(1) For the problem of the unknown model in stochastic system, the two control strategy of identification control and dual control is presented.The algorithm idea first to identify and second to control is used in identification control, identifying the unknown parameters through least square (LS), then control system by minimum variance control; The algorithm idea identification and control in one circle is used in dual control, estimating parameters by kalman filter. Finally compared the control effect of two methods in the process of system operation parameter mutation control by Matlab simulation, come to the conclusion the strategy between identification and control are feasible and effective in dual control.(2) For the stochastic system is composed of multiple model, based on the LQG control scheme theory, identifying the real parameter model from multiple models via joining the posterior probability, and when the real model will switch, the algorithm can identify the true parameters models, finally the feasibility of the presented algorithm is verified.(3) An efficient dual control scheme is presented for the stochastic parameters in dual uncertainty system. The control goal from single target to multiple target about recognition and tracking by joining identification on the basis of the traditional tracking target, it is convenient for the system to adjust the identification and tracking effect by different weights, Matlab simulation results show that this method between the parameter identification and target tracking with good control effect.(4) For the turn off in dual control process for stochastic, an improved controller is designed through setting threshold to the controller.Based on dual control in stochastic system, the controller may become zero as the change of covariance matrix, it will happen turn-off phenomenon, when strategy less than the threshold value through joining a threshold to the controller in this paper, the controller will assign a new value to avoid the turn-off phenomenon. Finally, the simulation results show that the validity of the approach to perfect the phenomenon in this paper.