Sensor networks have many advantages such as low power, low cost, distributed and self-organizing, easy resource sharing, and so on. They are widely used in target tracking, environmental monitoring, traffic control, health care and other areas. During the sensor data exchanging over the network, there often exist network congestion, packet loss and other problems due to the limited communication bandwidths, energy-efficient constraints, and so on. Recently, it has been the hot issue how to allocate resources and to compensate the packet losses efficiently when the network resources are limited. For the network constraint and packet dropout problem in sensor networks, this paper studies the information fusion estimation algorithms for sensor networked systems with network constraints and packet dropouts by using optimization theory and information fusion estimation theory. The main contents are as follows:The distributed fusion filtering problem is studied for dynamic stochastic variable with bandwidth or energy-efficient constraints. In sensor networks, each sensor gives its local filter based on its own measurement data. Due to the limited bandwidth, the quantized local filter is transmitted to the fusion center. In the fusion center, the distributed fusion optimal linear unbiased filter is designed based on the quantized local filter. The bandwidth scheduling is given to minimize the total transport energy consumption under the limited bandwidth constraint. Further, the approximate solution for the optimization problem is given under a limited bandwidth constraint.The weighted measurement fusion and distributed fusion estimation problems are studied for complex sensor networked systems with missing measurements, state and measurement multiplicative noises and transmission noises. A group of Bernoulli distributed random variables are used to describe missing measurements. Different sensors have different missing measurement rates. Based on full-rank decomposition of a matrix and weighted least-squares theory, the weighted measurement fusion estimators are developed by transferring multiplicative noises to additive noises. The weighted measurement fusion estimators have the same accuracy as the centralized fusion estimators, i.e., they have the global optimality. Also, for each sensor subsystem, the local estimators and the estimation error cross-covariance matrices between any two sensor subsystems are derived. Then, the distributed fusion estimators weighted by matrices in the linear minimum variance sense are given.The optimal fusion estimation problem is studied for sensor networked systems with random packet dropout compensations. A group of Bernoulli distributed random variables is employed to depict the phenomena of randomly packet loss in data transmission from sensors to estimators, and one step prediction value of the state is used as a packet loss compensation value. By applying completing method, the local optimal linear estimators including filter, predictor and smoother are given in the linear unbiased minimum variance sense. Further, the distributed optimal fusion estimators are given by applying the fusion algorithm weighted by matrices in the linear minimum variance sense. The cross-covariance matrices are derived between any local estimation errors. At last, the centralized fusion estimators are given. The accuracy comparison among them is simulated.The weighted measurement fusion quantized estimation problem is studied for sensor networked systems with limited bandwidth constraints and packet dropouts. There exist the phenomena of packet losses due to bandwidth constraints during the transmission. A Bernoulli distributed random variable is introduced to describe the phenomena of randomly packet loss. Based on the quantized measurement data received by the fusion center, two weighted measurement fusion quantized filters are presented. One is dependent on the value of the Bernoulli random variables. The other is dependent on the probability of Bernoulli random variables. They have the reduced computational cost and same accuracy as the corresponding centralized fusion filter. Also, the approximate optimal solution for the optimal bandwidth-scheduling problem is given under a limited bandwidth constraint. |