An Integrated System Using Federated Kalman Filter for UGV Navigation in GNSS-denied Environment

In global navigation satellite system (GNSS) denied areas such as urban canyons, accurate positioning for unmanned ground vehicles (UGV) is a challenging issue. This paper presents a novel method of an integrated navigation system which combines the strapdown inertial navigation system (SINS), GNSS, and the vehicle kinematics model. The integrated navigation system is built using a federated Kalman filter (FKF). For coping with abnormal GNSS signals, we make an adaptive modification of the classical Kalman filter using the proposed adaptive factor. In addition, the vehicle model is combined with four-channel wheel speed information and vehicle steering information to make up for the deficiency of information in traditional kinematics constraints. In order to maintain the high accuracy of the system, we propose a method for adaptively determining information sharing coefficients, which can assign weights to both local filters according to their own confidences. The experiments conducted in urban areas show that the new integrated navigation strategy can guarantee meter-level positioning accuracy within 130 seconds in GNSS-denied environments.