Distributed Filtering For Multi-Rate Nonlinear Systems Via Sensor Networks

As we all know,the sensor network is composed of a large number of sensor nodes with a certain topology.Sensor networks are broadly used in many real world applications,such as agricultural product quality inspection,home automation,health monitoring,intelligent transportation and so on.Sensor nodes can collect and transmit data,so it has attracted the attention of many researchers.On the one hand,the information can be exchanged among sensor nodes,and the distributed filtering method is more advantageous in terms of reliability and cost than the traditional centralized filtering method.On the other hand,in practical industrial production,the system is usually nonlinear,and most of the signals of different system components are sampled at different rates in order to save resources and ensure system performance.Therefore,it is a very significant project to propose some distributed filtering methods for multi-rate nonlinear systems.In this thesis,the distributed filtering algorithms under variance constraint index and boundedness discriminatory conditions are presented for several classes of time-varying multi-rate nonlinear systems.Specific studies are given as follows.(1)The resilient distributed filtering problem based on memory-event-triggered communication mechanism is studied for time-varying multi-rate nonlinear systems with integral measurements.The iterative method is utilized during the analysis of the multi-rate system and the phenomenon of integral measurements is considered to conform to practical engineering.To ease the communication burden,the memory-event-triggered communication mechanism is adopted,which decides whether the sensor nodes communicate with each other at measurement sampling instants.To improve the robustness of the filtering algorithm,the stochastic gain fluctuations are considered.The resilient distributed filter is constructed based on the information received.Next,an upper bound on the filtering error covariance is obtained and the suitable filter gains are selected to minimize it.In addition,a sufficient condition is proposed,which assures the uniformly boundedness of the upper bound on the filtering error covariance.The accuracy of the proposed distributed filtering algorithm is tested and the effect of the weighting coefficient on the triggered frequency is shown by a numerical example and a comparative experiment.(2)The outlier-resistant distributed filtering problem based on the strategy of amplify-and-forward relays is studied for time-varying multi-rate nonlinear systems with multi-measurement delays.To lengthen the signal transmission distance and ensure the communication transmission quality,we set an amplify-and-forward relay between the sensor and the filter.We modify the multi-rate system to a single-rate system by the augmentation method.Considering the phenomenon of measurement outliers,the outlier-resistant distributed filter is developed.Next,the upper bound on the filter error covariance is attained and then minimized by designing the filter gain.A sufficient condition is proposed,which bounds the upper bound on the filtering error covariance.A numerical example is used to illustrate the validity of the proposed distributed filtering algorithm and its robustness about measurement outliers.(3)Based on encoding-decoding scheme,the distributed fusion filtering problem is studied for time-varying multi-rate nonlinear systems with sensor resolutions.The iterative method is applied in the conversion process of the multi-rate systems and the sensor resolution is considered with the aim to closely match the actual engineering.An encoding-decoding scheme based on dynamic quantization is introduced to improve the efficiency of transmission and ensure the security of the source.The local distributed filter is constructed according to measurements after decoding.It is followed by deducing an upper bound on the covariance of the local filtering error and deducing the local filtering gains by minimizing the upper bound’s trace.Subsequently,on the basis of the covariance intersection fusion criterion,the fusion filtering algorithm is obtained.Moreover,a sufficient condition is introduced,which provides the uniformly bounded upper bound on the local filtering error covariance.A numerical example and a practical example of moving target tracking are used to illustrate the superiority of the proposed algorithm and discuss the monotonicity of the mean square error of the proposed fusion filter about the sensor resolution.