A zero-velocity update method based on neural network and Kalman filter for vehicle-mounted inertial navigation system

Abstract

Zero-velocity update (ZUPT) is an effective method of restraining the error divergence of the inertial navigation system (INS). The correct detection of zero-velocity points and an appropriate filtering algorithm are the key factors for the success of ZUPT. In this paper, a ZUPT method for vehicle-mounted INS based on a neural network (NN) and Kalman filter is proposed. The efficiency and accuracy of the zero-velocity detection is improved by the NN. The precision of the proposed method can reach 99.19%, and the recall rate is improved by 24% compared with the method based on the support vector machine. In addition, this method has similar accuracy and better real-time performance than the method based on a long short-term memory. Based on the zero-velocity detection by the NN, the navigation error is estimated and compensated by the Kalman filter. The effectiveness of the proposed method is proved by a vehicular experiment that shows that the velocity error is reduced to 24.2% and the position error is reduced to 9.5%.