One-Step Prediction-Enhanced FIR Filter and its Application in INS/Vision-Integrated Mobile Robot Localization

Abstract

In this article, to address the problem that some inertial navigation system (INS) data are not used for data fusion, which affects the accuracy of the system with INS and vision, one-step prediction-enhanced finite impulse response (FIR) filter will be introduced in this work. First, the integrated scheme for fusing the INS technique with vision technology is present. Then, the FIR filter with iterative form will be derived based on the INS/vision-integrated data fusion model. Third, the one-step prediction will be employed by the FIR filter for improving the localization error. The validity of the method derived in this work has been verified by a practical experiment. Test results demonstrate that the FIR algorithm can enhance the localization accuracy when compared with the Kalman filter, respectively, by employing the one-step prediction. Also, the proposed method shows more robust when compared with the Kalman filter.