Study On Finite-Time Control For Nonlinear Systems And Its Applications In TCP/AQM Networks

With the development of society and innovation of science and technology,the dynamic characteristics of the system have been known and understood more deeply.Meanwhile,the better system performance needs to be satisfied.In the existing control techniques,the finite-time control has been an important research field in recent years,which is also a main branch of the optimal control.It is hoped that the finite-time stability can be achieved in many practical applications.For example,the vehicle can stop as soon as possible owing to the emergency braking,the objective is fast captured in the manhunt problem,the missile rapidly hits an aircraft,and the artificial satellite can be launched to a designated space,etc.As a result,the study on finite-time control problem has not only theoretical significance,but also application value to some extent.It is well known that most of the real models are nonlinear,thus,nonlinear systems become a major research object in this dissertation.Via reading and summarizing the existing work,the design issue with respect to the finite-time controller and prescribed finite-time controller for several kinds of nonlinear systems,such as strict-feedback,pure-feedback,nonstrict-feedback,is considered in this dissertation,by using backstepping method,neural networks,fuzzy logic systems,H? theory,adding a power integrator(AAPI)and dynamic surface control(DSC)to name just a few.Nowadays,Internet is rapidly developing and the number of users is sharp increasing.In order to ensure people's living quality and work efficiency,how to avoid the network congestion and promote network performance has been a burning trouble.Up to now,the end-to-end transport control protocol(TCP)is still widely used in the Internet.For the sake of guaranteeing,the convergence of data transport queue to a desired value,the reason why the active queue management(AQM)algorithm is utilized instead of the passive queue management(PQM)is that the PQM algorithm is easy to generate deadlock,full queue and global synchronization.Reasonable allocation of network resources,congestion avoidance as well as stable and reliable network operation can be realized by exploring TCP/AQM network systems.In this dissertation,congestion control means that the possibility of the network congestion can be reduced to the minimum with the aid of taking certain measures or the network system can recover to the normally running state after the congestion appears,which coincides with the idea of finite-time control.Therefore,how to apply the finite-time control or prescribed finite-time control scheme of nonlinear systems to TCP/AQM network systems becomes another chief research content.The main contents are stated as follows.The research history on the finite-time control and network congestion control is introduced.Meanwhile,some basic concepts,lemmas,mathematical models and inequalities are introduced.The finite-time H? control problem is investigated for a class of strict-feedback nonlinear systems with external disturbances.Based on AAPI approach,backstepping technique and H? control theory,a sufficient condition of the finite-time stability is obtained and the corresponding design process of controllers is given.Moreover,the presented method is applied to TCP/AQM networks.Simulation results illustrate the effectiveness and feasibility of the proposed scheme.The semi-globally practical finite-time control issue is studied for a group of uncertain nonlinear systems.By virtue of combining AAPI and DSC methods,a lesscomplexity design process is given and a sufficient condition of the semi-globally practical finite-time stable is achieved.The proposed scheme is verified to be effective by two simulation examples,that is,a numerical example and a TCP/AQM application.Under the condition of the output constraint,the adaptive fuzzy and neural networks prescribed finite-time control problems for a kind of nonstrict-feedback nonlinear systems are addressed.Inspired by the idea of the prescribed performance control(PPC),a new performance function,called prescribed-time performance function(PTPF),is proposed for the first time.Furthermore,an explicit expression on PTPF is given.Meanwhile,another new concept,called practically pre-assigned finite-time stability,is defined.In the following,a novel sufficient condition of the finite-time stability is shown by combining the adaptive control,fuzzy and neural network approximation techniques,and backstepping method.Finally,the simulation results,including two numerical examples,a TCP/AQM network and an electromechanical system,are used to demonstrate the feasibility of the presented method.The adaptive fuzzy finite-time controller design problem for a kind of strict-feedback systems with the state constraints is considered.First,another novel PTPF is defined,which can be adjusted more flexible.After that,a non-affine system is established by using the coordinate transformation.Moreover,the sufficient condition of the practically finite-time stability is achieved by means of constructing the non-affine variable.Finally,two simulation results are given to illustrate the feasibility and effectiveness of the presented scheme.Design problem of adaptive prescribed finite-time controller for pure-feedback systems with full state constraints is discussed.Meanwhile,a series of modified barrier Lyapunov functions(MBLFs)are established for the first time.Next,with the help of mean-value theorem and MBLFs,all the states not only satisfy the given state constraints,but also pre-assigned steady and transient state performances.In the simulation experiment,six cases are considered to explain that the settling time is independent of the initial conditions and design parameters.Finally,the presented scheme is applied to the congestion control of TCP/AQM networks.An adaptive neural network prescribed finite-time quantized problem of nonlinear systems with time-delays is solved.Based on the direct adaptive control and MBLFs,the proposed control target is achieved and the constructed approach of controllers is also given.The presented technique is applied to chemical reactor systems to intuitively declare the proposed scheme effective.The main work of this dissertation is summarized and the future research direction is indicated.