英文摘要 | A class of feedback control system which is through real-time communication network in order to achieve the information exchange, resource sharing between the various components (such as: sensors, controller, actuator) is known as network control system. Network control system has the following advantages:to share information resources, greatly reducing the number of links, ease of expansion, high efficiency, low cost, fault diagnosis ability to facilitate the installation and maintenance and so on. Due to network bandwidth constraints, resulting in data transmission delay, data loss, data error and data sequence a series of problems such as misinformation, which makes the network control system analysis and design has become very complicated. In this paper, we study network control system based on variable-period sampling model, taking the current moment delay as sampling period, as delay is random and the corresponding sampling period is random. By use of linear matrix inequalities, convex optimization technologies, stochastic optimal feedback control, iterative learning control and minimum entropy control, focus on the introduction of new control strategies. Based on variable-period sampling model for network control systems, we deduced linear matrix inequality conditions to guarantee the network control system stochastic stable. Using convex optimization techniques to solve the LMI, we gain the state feedback controller. Similarly, based on the variable-period sampling model, using stochastic dynamic programming knowledge, combining with Lyapunov functions, we derived the optimal state feedback control law to make the network control mean exponential system stable. Taking the entire network control system as a stochastic system, we introduce the concept of entropy into NCS and design PI controller to guarantee the entropy of tracking error minimum. We achieve PI controller gain using iterative learning control techniques. In the framework of stochastic systems, assuming that the probability density function of delay is known, we use BP neural network to approximate the probability density function of output tracking error and make the entropy of the output tracking error minimum by the steepest descent optimization method in order to minimize uncertainty. Using gray prediction theory to predict the random network delay, we take the prediction of random time delay as the current sampling period and model the controlled object into a variable sampl... |
修改评论