To guarantee the navigation accuracy of integrated MIMUs/GPS during the period of loss of GPS signal, the Adaboost optimized BP neural networks are introduced in the normal Kalman filter.When GPS signal is unavailable, the trained neural networks are adopted to predict the velocity difference & position difference between outputs of GPS & MIMUs at the same moments, which then will be transmitted into the observer of Kalman filter.Specifically, by improving the navigation strategy on the systematic level, the error of navigation parameters accumulated from MIMUs under single mode is corrected.Simulation results indicate that: 1) When losing the GPS signal, the Adaboost optimized BP neural network can help the filter for information processing and compensate for the lost information in 50 s; and 2) Compared with only applying the normal BP or RBF neural networks, better stability and precision of navigation parameter prediction are obtained.