Visual-inertial odometry using iterated cubature Kalman filter

In recent years, there are many excellent algorithms for visual-inertial odometry. However, in practical engineering application, we must consider the computational cost. High-precision optimization-based methods usually can not meet the requirement of real-time. In this paper, we proposed a visual-inertial odometry algorithm, which is based on iterated cubature Kalman filter. Compared with EKF-based method, it can reduce the influence of linearization and improve localization precision. The computational complexity of this method is similar with other Kaiman filtering based methods. Experimental results are presented for a real-world dataset captured on a Beijing street with a land vehicle and the results show that the method proposed can attain a better accuracy than other methods.