Information fusion method based on adaptive SSUKF for integrated navigation system

The unscented Kalman filter (UKF) is studied as a state estimation method for the nonlinear system and is used to train multilayered neural networks by augmenting the state with unknown connecting weights. Because the computational costs of UKF are proportional to the sigma points, the UKF can not meet the real time requirement in system state estimate. The neural network-aided adaptive spherical simplex unscented Kalman filter (SSUKF) for the vehicle INS/GPS integrated navigation system is studied. In this algorithm a multiplayer neural network is used to estimate the system noise. And the SSUKF is used to estimate the state vector of vehicle INS/GPS integrated navigation systems and online train the multilayer neural network. The theoretical procedure of this algorithm is described in detail. Then, this algorithm is used in integrated navigation system when the statistic of system noise is unknown. Simulation results prove the availability of this algorithm. Not only can surely estimate accuracy be obtained, which is similar to that of the neural network- aided adaptive UKF, but also the run time is reduced considerably.