Information Fusion Estimation For Stochastic Uncertain Systems With Finite-step Correlated Noises

With the expansion scale of modern control system,the increasing degree of decentralization and the rapid development of network technology,the networked control system is widely applied in military and national economy by virtues of its high efficiency,high reliability,operation flexibility and low cost,etc.However,in the same time,the unreliable network environment makes the networked control system undergoing many uncertainties.In addition to random phenomena such as transmission delays and packet dropouts caused by limited network bandwidth,there are also many unavoidable problems.For example,process disturbances,correlated noises,random uncertainties,random nonlinearities and multi-rate sampling which arising from complex environment or limited economic conditions.In view of the above features of the networked systems,comprehensive considering the random phenomena in the process of network transmission,this paper studies the fusion estimation problems for the networked systems based on the projective theory and the optimal fusion algorithm in the linear unbiased minimum variance sense.The main research contents are as follows:1.The centralized and distributed fusion estimation algorithms are proposed for multi-sensor stochastic uncertain systems with multiplicative noises and correlated additive noises.The multiplicative noises in the state equation and observation equation are cross-correlated at the same time stamp for the considered system model.The system noise and observation noises are one-step auto-correlated and two-step cross-correlated.The optimal centralized fusion estimators are proposed in the linear minimum variance sense via an innovation analysis approach,including filter,predictor and smoother.Based on the optimal local filters,the filtering error cross-covariance matrices between any two filters are deduced.Furthermore,a distributed fusion filter weighted by matrices is put forward based on the optimal matrix weighted fusion rules in the linear unbiased minimum variance sense.To avoid the calculation of estimation error cross-covariance matrices,a CI(Covariance Intersection)fusion estimation algorithm is also proposed.At last,the comparative analysis is conducted for proposed three fusion algorithms in terms of estimation accuracy and computational burden.2.The centralized and distributed fusion estimation algorithms are proposed for the multi-sensor stochastic uncertain systems with multiplicative noise and finite-step auto-and cross-correlated additive noises.Base on the measurement augmentation method,the optimal centralized fusion estimators are designed in the linear minimum variance sense,including filter,predictor,and smoother.The optimal local filters are derived for every single sensor system.The filtering error cross-covariance matrices between any two local filters are deduced based on the innovation analysis approach.Then a distributed fusion filter weighted by matrices is proposed based on the optimal matrix weighted fusion rules in the linear unbiased minimum variance sense.Finally,simulation examples are given to show the effectiveness of the proposed algorithms.3.The linear minimum variance optimal estimators are proposed for multi-sensor stochastic uncertain systems with random parameter matrices,random nonlinearities,fading measurements and finite-step auto-and cross-correlated additive noises.Multiplicative noises,transmission delay and packet drops can be described by or transformed into random parameter matrices.The random parameter matrices considered in the state equation and observation equations are cross-correlated at the same time stamp.The fading ratios of different channels are different.All random nonlinearities are described by statistical characteristics.The optimal linear estimators are proposed via an innovation analysis approach,including filter,predictor and smoother.The estimation accuracy of the proposed algorithm is better than existing literature.4.A distributed fusion filtering algorithm is derived for the multi-rate multi-sensor stochastic uncertain systems with multiplicative noises and correlated additive noises.The state of the systems considered is updated at the highest rate.The measurement sampling rates are positive integer multiples of the state updating rate,and different sensors sampling uniformly with different rates.The local optimal linear filters at measurement sampling points are given by establishing the new system model at measurement sampling points.Then the local optimal linear estimators at state updating points are derived by using multi-step predictor.The estimation error cross-covariance matrices between any two local sensors are deduced,including filtering error cross-covariance matrices,estimation error cross-covariance matrices between filtering and prediction,prediction error cross-covariance matrices.They can be recursively computed.A distributed fusion filter weighted by matrices is proposed based on the distributed weighted matrix fusion estimation algorithm in the linear unbiased minimum variance sense.Simulation examples verify the effectiveness of the proposed algorithms.