【作者】 王勇；

【导师】 李昇平；

【作者基本信息】 汕头大学 ， 机械电子工程， 2008， 硕士

【摘要】 对网络控制系统分别从高斯型二次最优(LQG),切换系统和神经网络控制三个方面来考虑。1.用LQG方法考虑网络控制系统的丢包问题,仅考虑丢包发生在传感器到控制器之间。在传感器一边放置一个Kalman滤波器消除系统扰动和量测噪声,控制器一边放置一个系统模型。将系统建模为随机丢包时,利用马尔可夫链跳变系统的已有的结论给出系统稳定性结论。最后给出网络控制系统性能的充要条件。2.研究了当传感器和控制器出现丢包时的网络控制系统并将其建模为切换系统。因此网络控制系统的稳定性可以相应的转换为切换系统的稳定性来研究。用平均驻留时间的方法来分析周期性切换和随机切换。最后讨论当系统由稳定和不稳定的子系统组成时整个系统的L2增益要小于一个正数γ0。3.首先对网络控制系统进行建模。接着用BP神经网络对网络上丢失的数据包进行预测,采用Levenberg Marquardt(LM)方法来改进BP算法。我们只考虑数据包丢失仅发生在传感器到控制器之间。更多还原

【Abstract】 We consider the networked control system (NCS) from linear quadratic Gaussian (LQG) method, switched system and neural network control.1. We consider the problem of optimal LQG control of a system in which packets drop happened between the sensor and the actuator. We do not assume the packets drop behavior and achieve the optimal algorithm. The controller uses an explicit model of the plant via the packets drop happened. We place a Kalman filter at the sensor side to erase the process disturbance and measurement noise. We then use a stochastic packet-drop model and apply the result of a markovian jumping parameters to the system. Conditions for stability of the NCS are given. We also derive the necessary and sufficient condition for the performance of the NCS.2. We will model the NCS as a switched system when packets drop happened between sensor and controller which are connected by network. The stability of the NCS can be reduced to the corresponding problems for switched system. We will formulate the system as periodically switching and a stochastically switching incorporated with average dwell time method. Then we will study L2 gain problem that the whole system achieves L2 gain smaller than a positive scalarγ0 in the cases when the system are composed of stable and unstable subsystems.3. we will first analyze the modeling scheme of the NCS. And then use BP neural network to predict lost data when packets drop happened between plant and controller which are connected by network. We will use Levenberg Marquardt (LM) method which is an improvement of the BP algorithm.更多还原