EM-IMM based land-vehicle navigation with GPS/INS

Integration of the global positioning system (GPS) with the inertial navigation system (INS) is favorable since it provides enhanced positioning accuracy. Its implementation is essentially based on the standard Kalman filter techniques. However, the estimation accuracy is degraded if unknown parameters present in the system model or the model changes with the environment as in the case of intelligent transportation systems (ITS). We propose an expectation-maximization (EM) based interacting multiple model (IMM) method, namely, EM-IMM algorithm, to jointly identify the unknown parameters and to estimate the position information. The IMM is capable of identifying states in jumping dynamic models corresponding to various vehicle driving status, while the EM algorithm is employed to give the maximum likelihood (ML) estimates of the unknown parameters. Compared to the conventional single model Kalman filter based navigation, the proposed algorithm gives improved estimation performance when the land-vehicle drives with changing conditions. Simulation results demonstrate the effectiveness of the proposed method.