INS/VNS Fusion based on unscented particle filter
Dongxue Yue, Xinsheng Huang, Hong-Li Tan · 2007
In vision aided terminal inertial navigation system (INS), the INS use dead reckoning calculate the velocity and position of the vehicle based on the measurements of the strapdown inertial sensors. Using the parallel structure of the runway, vision navigation system (VNS) can estimate the relative position and velocity of vehicle to the runway, based on the error position of the feature points of the runway in perspective image and weak perspective image. For the fixed target point on the runway, the relationship of the velocities of INS and VNS are deduced, on which unscented particle filter (UPF) is designed to fuse the two sensors. The resulting output of the filter is the velocity error of the vehicle. When fixed bias exists in outputs of accelerometers or gyroscopes, the process noise is not Gaussian white noise any more. UPF can deal with any distribution and do not need increase states to model the noise as Kalman filter does. Simulation results show that the UPF fusion algorithm can increase the accuracy of the velocity and improve the performance of the navigation.