On The Estimation Algorithms For Networked Multi-Sensor Fusion Systems

Multi-sensor information fusion, which has attracted people’s great attention, is a ris-ing multi-discipline field started from the1970s with the urgent need of the development of military affairs, national defense, wars and high-tech. As one of important issues in in-formation fusion, the multi-sensor fusion estimation problem has been a focus of research because of their wide application in military and civilian fields. With the rapid development of network communication technology, communication network is introduced to connect the distributed sensors and FC, and this class of systems may be called networked multi-sensor fusion systems (NMFSs). Compared with classical multi-sensor fusion systems, the insertion of the communication network in NMFSs can offer many advantages such as flex-ible architectures, simpler installation, easier maintenance and low cost. Therefore, NMFSs have now been one of the hot research topics.Although NMFSs have brought so many advantages, they also bring lots of new prob-lems and difficulties, thus the conventional fusion estimation approaches will not be ap-plicable to the NMFSs. In this case, an important and practical problem is how to design fusion estimation algorithm for the NMFSs. Under this background, based on the projec-tion theory, Lyapunov theory, matrix analysis theory and optimal weighted fusion algo-rithm, this paper is concerned with two important prolems existing in NMFSs:one is the communication bandwidth and sensor energy constraints, the other is the problem of un-certain measurement, random delay and packet dropouts. The main work is summarized as follows:1. The distributed fusion estimation problem is investigated for a class of NMFSs with communication bandwidth constraints. On one hand, a dimensionality reduction strategy with structural limit is proposed to satisfy finite bandwidth, then based on the optimal es-timation fusion algorithm weighted by matrices, a recursively distributed Kalman fusion estimator is derived, and a simple suboptimal judgement criterion is proposed to determine a group of binary variables such that the mean square error (MSE) of the designed estimator is minimum at each time step. On the other hand, when the process and measurement noises in the NMFSs have unknown statistic characteristic but bounded energy, a group of finite-level logarithmic quantizers are introduced to describe the case of bandwidth constraints. By using H∞filtering theory and the discrete-time bounded real lemma, the necessary and sufficient condition is derived such that the performance of the distributed H∞fusion es-timator is optimal. Under this condition, the optimal weighted matrices and quantization parameters are given. The target tracking system and an illustrative example are given to demonstrate the effectiveness of the proposed methods.2. The distributed Kalman fusion estimation problem is investigated for a class of NMFSs with bandwidth and energy constraints. To satisfy the finite communication band-width, at a particular time, only partial components of each local estimate are allowed to be transmitted to the FC in a random way, while each sensor intermittently sends information to the FC for reducing energy consumptions. Then a recursively distributed fusion Kalman estimator is derived in the linear minimum variance sense. Since the performance of the de-signed estimator is dependent on the selecting probability of each component, some criteria for the choice of probabilities are derived such that the MSEs of the designed estimators are bounded or convergent. The steady-state distributed fusion Kalman estimator is also given. Finally, the target tracking system is given to demonstrate the effectiveness of the proposed methods.3. The distributed mixed H2/H∞fusion estimation problem is investigated for a class of NMFSs with limited communication capacity, where the system perturbations are mod-eled by white noise and bounded energy noise. The stochastic dimensionality reduction strategy with structural limit and logarithmic qunatization strategy are simultaneously taken into account for deducing the traffic between the distributed sensors and FC. By resorting to Lyapunov theory and the mixed H2/H∞filtering approach, a sufficient condition, which is dependent on the quantization parameters and the selecting probabilities of the trans-mitted component, is derived such that the distributed mixed H2/H∞fusion estimator is stable. For given the transmitting probabilities of the estimate component, the design ap-proach for the optimal weighted matrices and quantization parameters is presented under the communication capacity constraints. Finally, the F-404aircraft engine model is given to demonstrate the effectiveness of the proposed methods.4. The fusion estimation problem is investigated for a class of NMFSs with sensor fail-ures, stochastic parameter uncertainties, random observation delays and packet dropouts. A novel model is proposed to describe the random observation delays and packet dropouts, and a robust optimal fusion estimator is designed by using the innovation analysis method and algebraic Riccati equation. The dimension of the designed estimator is the same as the original system, which helps reduce computation cost as compared with the augmentation method, thus the designed estimator can satisfy the real-time performance of the system. Moreover, robust reduced-dimension observation fusion Kalman estimators are proposed to further reduce the computation burden. Some sufficient conditions for stability and optimal-ity of the designed fusion estimators are given, and steady-state Kalman fusion estimator is also presented. Simulation results show the effectiveness of the proposed methods.5. The distributed Kalman fusion estimation problem is investigated for a class of NMFSs with missing sensor measurements, random transmission delays and packet dropouts. A novel stochastic model is proposed to describe the transmission delays and packet dropouts, and an optimal distributed Kalman fusion estimator is designed based on the optimal fusion criterion weighted by matrices. Some sufficient conditions are derived such that the MSE of the designed estimator is bounded or convergent. Moreover, steady-state distributed Kalman fusion estimator is also presented. Finally, the target tracking system is given to demonstrate the effectiveness of the proposed methods.