Research On Flnite-Time Intelligent Tracking Control Algorithms Of Pure-Feedback Nonlinear Systems Based On Neural Networks

In the 21 st century,with the rapid development of science and technology as well as society,there is a strong intersection between various disciplines.At present,the complexity and coupling of engineering application systems have increased significantly.Therefore,the research on intelligent control of pure-feedback nonlinear systems based on modern control theory has more and more extensive application prospects.This thesis focuses on the important and interdisciplinary scientific problem of “finite time intelligent tracking control of pure feedback nonlinear systems based on neural network”.With the continuous improvement of the performance requirements of control system in modern industrial production,the control objectives in practical engineering are often expected to be achieved in finite time.On the premise of the above analysis,this thesis presents the finite/fixed time and predefined time intelligent tracking control algorithm for a class of pure-feedback nonlinear systems respectively.Furthermore,the adaptive tracking control problem of pure-feedback nonlinear multi-agent systems(MASs)is solved and an intelligent tracking control algorithm based on special-shaped Laplace matrix design method is proposed.The main contents of this thesis are as follows:(1)The practical finite/fixed time intelligent tracking control algorithm is proposed for the pure-feedback nonlinear systems which nonaffine functions may be nondifferentiable.Firstly,a novel decoupling technique is used to eliminate the limitation of the partial derivatives of non-affine functions.Secondly,the practical finite/fixed time stability framework for the pure-feedback nonlinear system is established and applied to practical flexible robot system.In addition,in order to save communication resources more effectively,the relative threshold event triggering mechanism(ETM)is cleverly introduced in the transmission channel from the original control signal to the actual control signal.The tracking error converges to an arbitrarily small set near zero in finite/fixed time.Finally,the feasibility of the proposed algorithm is verified by two simulation examples.(2)The predefined time intelligent tracking control algorithm based on a novel switching threshold ETM is proposed for a class of pure-feedback nonlinear systems.Firstly,a new event triggering method is proposed in this thesis.Different from the existing research results,the new ETM proposed in this thesis can better avoid excessive pulse when the controller amplitude is large.Secondly,a practical predefined time and predefined performance stability framework is established for the considered system.The problem of “complexity explosion” in the backstepping process is solved by combining finite time differentiator with radial basis function neural networks(RBF NNs)technology.The predefined time intelligent tracking controller can not only stabilize the system state within the predefined time,but also ensure the tracking error reach the predefined accuracy within the predefined time.Finally,the effectiveness of the algorithm is verified by a simulation example.(3)A special-shaped Laplace matrix design method is proposed for a class of pure-feedback nonlinear MASs.Furthermore,an intelligent tracking control algorithm is proposed based on this design method.Firstly,an adaptive stability framework is established for a general pure-feedback nonlinear agent system and is applied to the considered MASs.Secondly,in the design framework based on Backstepping,the special properties of RBF NNs are used to filter out the extra signals in the design process.At the same time,ETM is introduced to reduce the transmission frequency of control signals.In addition,a special-shaped Laplacian matrix method is proposed for MASs,which solves the problem that the form of leader gain has not been unified in the process of tracking error design of MASs.According to the proposed cooperative adaptive stability framework,it is proved that all the closed-loop signals of the pure-feedback nonlinear MASs are bounded and the tracking errors can converge to an arbitrarily small neighborhood near the zero point.Finally,the effectiveness of the proposed algorithm is verified by a simulation example.