Research On Networked Predictive Control Of Affine Nonlinear Systems And Bilinear Systems

Network Control Systems (NCSs) are a class of feedback control systems that e-merged with the widespread use of computer network and the development of network technology. System components, like sensors, controllers and actuators, connected with the controlled plant via limited bandwidth digital communication networks. Thus it makes remote control and resource sharing come true. However, the insertion of the communication network will inevitably lead to network-induced delay, data pack-et dropout and packet disorder, which usually resulting in performance degradation or even instability of systems. To deal with these problems, Guoping Liu, David Rees et al. introduced a networked predictive control (NPC) scheme combining control theory and the characteristics of network data transmission. The system states and control inputs transported from the sensor will be used to calculate a predictive control sequence by the Control Prediction Generator (CPG) that located on the controller. Then the control predictions are packed and sent to the plant side through the forward network channel. Next, the Network Delay Compensator (NDC), located on the actuator side, chooses the appropriate control input from the latest prediction sequence. Thus, the networked issues will be compensated actively based on the NPC scheme. In this dissertation, the networked predictive control theory will be extended to two kinds of nonlinear systems: affine nonlinear system and bilinear system, initially. The main research work is shown as follows:(1) Predictive control-based approach and stochastic stability analysis to networked affine nonlinear systems are investigated. Firstly, the affine nonlinear system is lin-earized according to the control law that based on the system characteristics. Then the networked predictive control scheme is applied to the linearized system, in which the round-trip delay (both in feedback channel and forward channel) is modelled as Markov chains. After exchanging the original system to an augmented system, the equivalence sufficient stability conditions are obtained based on the Lyapunov stability theory and the definition of stochastic stability. Finally, a numerical example is given to demon-strate the effectiveness of the proposed method.(2) Predictive control-based approach and globally asymptotic stability to net-worked bilinear systems are investigated. Firstly, the one-step ahead state prediction is given according to the model of bilinear systems, and then the future predictive states are deduced. Based on these, a nonlinear predictive controller is proposed and the con-troller parameter selection method is given. Then the globally asymptotically stability is proved by selecting a scalar positive definite function. Finally, a comparison between the nonlinear predictive controller and the state feedback predictive controller is giv-en in the numerical example. The results of simulation demonstrate the effectiveness of the proposed predictive control scheme and the superiority of nonlinear predictive controller.(3) Predictive control-based approach to networked bilinear systems is investigat-ed and two algorithms of solving predictive control sequence are proposed. Firstly, the predictive states are obtained based on the situations that time-delay exists in the feedback channel and forward channel, respectively. Then the solving process of pre-dictive control sequence is converted to solve a non-convex problem, for which two gradually-optimized algorithms are proposed based on the special structure of bilin-ear system dynamics model. Algorithm 5.1, called the Stepwise Iterative Algorithm, shows the control sequence initial values firstly. Then the predictive control values will be iteratively calculated until the variation of performance satisfies the given threshold; Algorithm 5.2, called the Approximate Forward Stagewise Algorithm, will increase the dimensions of control vector and state vector by each step until all the approximate op-timal solution are obtained. Simulation results show that both the predictive control sequences calculated by the two algorithms can ensure good control performance.