| In the recent years,wireless sensor networks(WSNs)are widely used in the field of environmental monitoring,target tracking,healthcare applications etc.because of their powerful data acquisition and parallel signal processing abilities.WSNs are composed of a large number of sensors which cooperate with each other to accomplish several complicating tasks such as tracking and estimation.WSNs own many characteristics such as large-scale,self-organization,dynamic topology,as well as own many constraints such as poor energy,limited computing capacity,limited and time-varying network bandwidth.A fundamental problem in the application of WSNs is to deal with detection,estimation and fusion problems using scalable algorithms.As an effective distributed data fusion method,the distributed Kalman consensus filter(DKF)does not require a fusion center.Each node only exchanges information with its neighbor nodes,and all nodes can achieve consistent with considerable accurate estimates.Aiming at the characteristic of time-varying network topology in WSNs,several types of distributed Kalman consensus filtering algorithms were designed in this thesis combined with three key performance indicators named fault tolerance,convergence speed and network lifetime.The main research work is summarized as the following three aspects:(1)A distributed Kalman consensus filter was designed under two classes of packet dropout.The Bernoulli binary random variables were used to describe packet loss happen or not firstly.And then the coupling relation between the observation packet loss and the communication packet loss was discussed.By matrix theoretic,the stability and filtering performance of the distributed Kalman consensus filter were analyzed.Finally,a simulation example was given to verify the results obtained by algorithm.(2)A distributed Kalman consensus filter with continuous packet loss compensation was designed.If one-step or continuous communication packet loss did happen,the latest packet was used compensating the system.As a result,the negative impact of packet loss on filtering accuracy was weakened.Moreover,the convergence speed of the algorithm was also accelerated at the same time.In addition,the optimal filtering gain was obtained by minimizing the estimation error covariance.Finally,a simulation example was given to verify the effectiveness of the proposed algorithm.(3)Two distributed Kalman consensus filters with adjustable energy were designed.Considering the characteristic of sensor node redundancy layout,two methods were used to regulate the network topology respectively in order to decrease energy consumption.In the first algorithm,the probability of active packet loss was adjusted based on the percolation model,to lessen the number of transmission packets so that obtain longer network lifetime at the cost of reducing filtering accuracy slightly.The second algorithm was designed based on the event-driven mechanism to determine whether the nodes communicate or not for cut down unnecessary data transmission.As a result,the algorithm also achieved the required accuracy with lower energy consumption.Finally,a simulation example was given to verify the effectiveness of the proposed two algorithms. |
