Vehicle Positioning And Orientation Data Fusion Technology Research

In urban environments, the positioning performance of the on-board GPS is notsatisfactory because of satellite signal being blocked and multipath effect. For this, theresearch group of this paper proposed a RFID based Active Vehicle Positioning (RAVP)method. The instantaneous positioning accuracy of this method is much higher thannormal GPS receivers. However, if using RAVP separately, the positioning error isgeneral proportional to the distance of two RFID tags. That means we need deploy alarge number of RFID tags of high density if we want to fulfill high accuracypositioning. So it has certain difficulty in performing RAVP in recent time.For this, we proposed a positioning method combining RFID and GPS in this paperto improve GPS positioning accuracy and reduce the number of the RFID tags. Thecontents of this paper including:1) Combine the Extended Kalman Filtering (EKF) and RAVP, using the precisecoordinates provided by RFID tags to correct the state estimation of the EKF.In addition, this method can continue working when the valid satellite numberis less than4and restrain the accumulation of the positioning error.2) Next, according to the RFID tags’ characteristic of precise positioning, theon-board GPS and RFID tags together form a virtue differential array. Thissystem can estimate the common error of GPS measurements like the normalDGPS. And it can restrain the drawback of uncommon error increasing in thenormal DGPS. Meantime, this method can estimate the observation noisevariance. The simulation results show that this method performs well inWeighted Least Squares (WLS) and EKF.3) After acquiring observation noise in2), we can use the observation noise inSage-Husa adaptive filtering algorithm to estimate system noise, which canguarantee Kalman Filter work well and avoiding filtering divergence.Meantime, for in the Sage-Husa algorithm the Q matrix can not be kept positivesemidefinite, we discuss a bias estimation method to keep the Q matrix positivesemidefinite and point out that the Q estimation is larger than the real value, but it still has some value in practical application. In addition, Zhang Changyunfrom Beijing University of Aeronautics and Astronautics proposed aproportional method to estimate Q value and he think through iteration hisalgorithm can approach the real Q value. In this paper, we demonstrate that theresult given by proportional method is lower than the real value. In the end, weproposed a new exponent method which can get an unbiased Q estimation. Thesimulation results shows that this method perform well.