Neuro-fuzzy adaptive Kalman filtering for INS/GPS integration

Adaptive Kalman filtering is widely used in data fusion and integration. This paper highlights the use of fuzzy inference system in INS/GPS integration for adaptation of the measurement noise covariance matrix, while the process covariance matrix is known. Selection of initial measurement noise covariance matrix, training process and an innovative adaption approach are investigated. A strapdown INS equipped with high quality inertial sensors and an accurate GPS receiver is evaluated in several tests, and the results confirm that tuning the measurement noise covariance matrix in neuro-fuzzy adaptive Kalman filtering approach outperforms the conventional Kalman filter in INS/GPS integration.