Multi-model self-tuning weighted fusion Kalman filter

For the multisensor single channel autoregressive moving average (ARMA) signal with a white measurement noise and autoregressive (AR) colored measurement noises as common disturbance noises, when the model parameters and noise statistics are partially unknown, a self-tuning weighted fusion Kalman filter is presented based on classical Kalman filter method. The local estimates are obtained by applying the recursive instrumental variable (RIV) and correlation method. The fused estimates are obtained by taking the average of all corresponding local estimates. Then the optimal weighted fusion Kalman filter is obtained by substituting all the fusion estimates into the corresponding optimal Kalman filter. A simulation example shows its effectiveness.