Research On Inverse Control Of Time-Varying Systems Based On Adaptive Filtering Algorithms

In practical applications, control systems with time-varying characteristics are very common. Therefore, the control schemes on this class of time-varying plant is more and more concerned in control field, and becomes one of the hot problems under discussing. Methods of modern control theory and some intelligence control schemes are usually designed and modified for a class of certain systems, and have a little consideration about time-varying characteristics. Moreover, related theories are necessary needing complemented. In this thesis, we plan to use adaptive inverse control method to solve the control problem on time-varying systems. In this scheme, methods of adaptive signal processing are borrowed from the field of digital signal processing to solve problems in dynamic systems control. Meanwhile, it also uses the inverse of plant model as a controller of the open-loop system. All of these characteristics are very valuable for time-varying systems. Compared with adaptive control method, adaptive inverse control can track the dynamic changes caused by plant, and decrease the dynamic noise caused by the whole system without excessive restrictions for controller and plant. Recently, adaptive inverse control for discrete systems are usually designed based on adaptive filters, therefore, this thesis also use this scheme, by comprehensive analyzing on the theory of adaptive inverse, proposed series of adaptive filtering algorithms and inverse control method for time-varying systems. The main contents and results are as follows:1) A brief review is made on development and research status quo of the control systems with time-varying characteristics, and the analyzing is proposed on specific character and application domains of the the time-varying control methods.2) Series of novel convex combination filtering algorithms are proposed in this paper. As traditional adaptive filtering algorithms are usually compromised by convergence speed and steady-state accuracy, the optimal performance can not achieve in this state. Meanwhile, the mean idea about the theory of convex combination adaptive filtering is convex combination the filters with difference convergence characteristics, so that the whole adaptive filter can play their advantages in different stage of convergence process and avoid the compromise caused by traditional filters. Based on the basic theory of combination filtering, some improved algorithms are proposed in this paper, on variable step size and low level of computational complexity respectively. Then, the performances are analyzed. Moreover, for time-varying systems with structural uncertain, the multi-combination filtering algorithm with different tap lengths are proposed, and the simulation examples show the novel methods have more effective performance.3) Series of improved particle swarm optimization algorithms are presented for designing IIR digital filters. IIR filter can solve the problem, which is caused by FIR filter, that the correlation matrix eigenvalues becoming large irregularly when it identifies the time-varying system. Meanwhile, IIR filter is able to reduce the weight vector length during the online training processing and to improve the efficiency of optimization and modeling. But it has only one shortcoming that the performance surface is unsmooth, has local minimum possibility. To solve this problem, the improved particle swarm optimization (PSO) algorithms are designed for IIR filters, which can provide a better method for global optimization and less computational complexity than the traditional standard PSO algorithms. Finally, the proposed PSO-IIR filters are well used in applications on time-varying plant modeling and identification.4) Adaptive filters designed in last two chapters are applied in linear time-varying adaptive inverse control system. Through building the structure of linear adaptive inverse control system, and analyzing the system requirements on adaptive filtering algorithms, the two kinds of adaptive filters (combination filter and IIR filter based on improved PSO algorithm) are used in applications of plant modeling and inverse plant modeling. Moreover, simulation experiments show that different adaptive filter algorithms have different effects on the performance of the whole inverse control system.5) For a class of nonlinear time-varying systems, an online adaptive inverse control method based on combination Volterra kernel filtering algorithm is proposed. This approach has ability to deal with the time-varying characteristics and present an online tracking control mechanism for complex dynamic systems. On this basis, novel combination Volterra kernel method is proposed for the nonlinear time-varying characteristics. Meanwhile, the new algorithm can also provide a better method to reduce computational complexity and realize a dynamic online tracking control mechanism for nonlinear systems than the original algorithms. Further more the simulation analysis shows that, for this type of complex time-varying systems, the adaptive inverse control method based on online combination filters can achieve online modeling of unknown plants fast and effectively, and also track the real-time changing in complex time-varying systems.Finally, the potential further research direction in the method of adaptive inverse control for time-varying systems is discussed after summarizing the whole work in this thesis.