Filtering And Fault Detection Of Networked Switched Positive Systems With Random Missing Measurements

Recently, switched positive systems, an important class of hybrid systems, have received an increasing research interest in virtue of their significance in both theory and applications. On another research field, as a new type of networked, intelligent and distributed control system, networked control systems meet the requirement which is increasing as more and more complex control tasks and the sharing and exchange of information from various components. However, at the same time, they also lead to some difficult problems, such as quantization, network induced delay and packet loss. In order to make full use of the advantages of networks and control a switched system effectively, it is usually required to design a filter and a controller which shares the communication channels. In addition, the effective fault detection measures need to be taken for avoiding the loss of the personnel and property caused by a fault. It is obviously significant to study networked switched positive systems. In this paper, we discuss the stabilization, filtering and fault detection of networked switched positive systems. The main work is summarized as follows:The stabilization problem of switched positive systems over a lossy network is studied. Firstly, we consider two cases, one is that data missing happens in the channel between the system and controllers, the other is that data missing happens in the channel between the system and observers. Data missing is governed by a bernoulli process. Secondly, sufficient conditions on the existence of controllers in both cases have been given, and the controllers and observers can be obtained by solving linear matrix inequalities (LMIs). Finally, the design method of the controller is given.The design problem of fault detection filter for switched positive systems is discussed, and the data missing occurs on the channel between the sensors and the filters. Suppose data missing is governed by a bernoulli process, a stochastic switched positive system is established based on the augmented systems of the plants and filters. Two criteria are developed to evaluate the per-formance of detecting the faults for such system. Sufficient conditions on the existence of filters for the mean stability with an L_ fault sensitivity level and an L1 disturbance attenuation level are derived through constructing a switched Lyapunov function in term of the strict linear matrix inequalities. The result is applied in Leslie matrix model, which is very useful for application in agriculture and forestry.The distributed filtering problem is investigated for a class of switched nonlinear positive systems over sensor networks. The randomly varying nonlinearities and missing measurements, which are governed by two mutually independent Bernoulli distributed white sequences, are taken into account. Based on the output measurements of the individual sensor and its neighbors, the distributed filter with positivity constraints is designed to ensure the prescribed average l^ performance index of the estimation error dynamics. Sufficient conditions on the existence of distributed filters are presented in terms of LMIs and the filter gains are constructed.