AR Signal Fusion Identification And Estimation With Incomplete Observation And Sensor Deviation

With the application of multi-sensor information fusion technology in many fields,the information fusion estimation theory has attracted more and more attention from domestic and foreign experts and scholars.As we all know,Kalman filtering,as a classic filtering method,has been widely used for signal denoising.However,the application conditions of Kalman filtering are relatively strict,and the system model cannot contain any unknown quantities,and must be accurately known.In engineering practice,system model parameters and noise statistical characteristics are often unknown,so some system identification methods must be used to identify unknown system model parameters and noise statistical characteristics to obtain adaptive or self-tuning estimation algorithms.Autoregressive(AR)signals can generally describe some actual signals.In an actual networked system,due to network limitation and other reasons,data transmission may have fade,loss,delay and other phenomena.Due to the sensor's own characteristics and aging problems,the data received by the sensor and the real data must exist bias.Therefore,it is of great theoretical and practical significance to study the problem of identification and estimation of multi-sensor AR signals based on incomplete measurement data and sensor bias.This paper studies the fusion identification and estimation of multi-sensor AR signals with incomplete measurement and sensor bias.The main contents are as follows:For a multi-sensor multi-channel AR signal system with unknown model parameters,unknown missing measurement rate,unknown sensor constant bias,and unknown measurement noise variance,the AR model is converted into a state space model.First,when the system is accurately known,a matrix-weighted distributed optimal fusion filter is given.Then when the system has unknown parameters,the multi-dimensional recursive augmented least squares(MRELS)algorithm is used to identify the AR model parameters and sensor bias,and the correlation function method is used to identify the measurement noise variances and receiving measurement rates.Finally,the identified estimates of the unknown parameters are substituted into the optimal filtering algorithm to obtain the corresponding self-tuning filtering algorithm,and prove that the self-tuning filtering algorithm converges to the optimal filtering algorithm.For the multi-sensor AR signal system with unknown model parameters,unknown statistical characteristics of fading measurement random variables,and random sensor biases,the AR model is transformed into a state space model,then augment the state and sensor bias to obtain an augmented equivalent state space model based on the state augmentation method,and give a distributed optimal fusion filter weighted by matrix when the system is accurately known.When there are unknown parameters in the system,the recursive augmented least squares(RELS)algorithm and the correlation function method are used to gradually identify the AR model parameters,the statistics of the fading measurement random variable,and the virtual measurement noise variance.Finally,the identified estimates of the unknown parameters are substituted into the optimal filtering algorithm to obtain the corresponding self-tuning filtering algorithm and the convergence of the algorithm is analyzed.For the multi-sensor AR signal system with unknown model parameters,unknown statistical characteristics of fading measurement random variables,and random sensor biases,the AR model is transformed into a state space model,and a new measurement is defined based on the measurement difference method,a new equivalent state space model is proposed.Different from the state augmentation method,at this time,the measurement noise of the new model is related to neighboring moments.When performing Kalman filtering,a measurement noise estimator is required.Matrix-weighted optimal and self-tuning distributed information fusion filters are proposed,and the convergence of the algorithm is analyzed.