Adaptation of model errors in Kalman filtering

Based on predicted residual analysis and estimation of filtering errors, a new Kalman filter approach to correct model errors is proposed. Considering the expression of predicted residual, there is no need to discuss whether model errors are from measurement model or dynamic model. Based on the estimation of filtering errors, we just need to correct the measurement model to clear up the whole effects of model errors on the filtering value at current time. A GPS kinematic navigation calculative experiment is given to prove the feasibility of this method. It not only eliminates the whole effect of Kalman filtering model errors efficiently, but also has a good result.