Research On Robust Fusion Filtering Method For Uncertain Networked Systems With Colored Noises

With the improvement of the level of network and communication technology,various systems in the field of automatic control are developing toward networking and intelligence.Networked control systems(NCSs)are more secure,efficient,and intelligent than the traditional direct control method.However,uncertainties in transmission,such as network transmission delays and packet dropout,inevitably occur due to the limited network bandwidth.Meanwhile,noises in the system are generally assumed to be white noises in the existing research,but the engineering practice system will be disturbed by colored noises.To address the above problems,this research applies the principle of minimax robust estimation(MRE)to investigate the robust Kalman fusion filtering problem for uncertain networked systems with colored noise,and the main research contents are as follows.Firstly,the Centralized Fusion(CF)algorithm is used to process three types of uncertain multisensor NCSs in which only the input noise is colored noise,the input noise and the observation noise are the same colored noises,and the input noise and the observation noise are different colored noises,respectively.The uncertainty phenomena in the communication transmission process are described by the Bernoulli random sequence.The system models are transformed by using augmentation,de-randomization,and fictitious noise techniques.Based on the principle of minimax robust estimation,the robust CF steady-state Kalman estimators(including predictor,filter,and smoother)are proposed in the sense of Unbiased Linear Minimum Variance(ULMV).The robustness of the proposed estimators is demonstrated by means of augment noises,non-negative definite matrix decomposition,and the Lyapunov equation.Secondly,a distributed fusion algorithm is adopted to address the two types of multisensor systems to avoid the problem of the large computational burden of the CF algorithm.For the uncertain NCSs where the input noise and the measured noise are linear correlated signals,a robust local steady-state Kalman estimator is designed first,and then the robust distributed fused steady-state estimator is obtained by a matrix-weighted fusion algorithm.For the uncertain NCSs with public interference,that is autoregressive(AR)colored noise,a robust distributed fused steady-state Kalman estimator is obtained by transforming the model through the state space and other methods,and then by using a weighted fusion algorithm by a diagonal matrix.The accuracy relationship between the proposed local and distributed fused estimators is analyzed.Finally,robust steady-state Kalman predictors are designed for multichannel AR signal uncertain NCSs with random parameter matrices.For the single-sensor observation case,the robust steady-state Kalman multi-step signal predictor is proposed by converting the signal model to a state-space model and by transforming the mixed uncertainty system to a single uncertainty form;then this single-sensor model is extended to a multi-sensor system and a robust matrix-weighted fused steady-state Kalman multi-step signal predictor with higher accuracy is designed.And the robustness of the designed signal predictor is rigorously demonstrated.The proposed method is simulated by Uninterruptible Power System(UPS)and springdamped system,and the simulation results verify the correctness and effectiveness of the proposed method.