A novel cubature Kalman filter for nonlinear state estimation

Abstract

The cubature Kalman filter (CKF) is more preferred over the unscented Kalman filter (UKF) for its more stable performance. The CKF employs a third-degree spherical-radial cubature rule to numerically compute the integrals encountered in nonlinear filtering problems. The third-degree cubature rule-based filter, however, is not accurate enough in many real-life applications. Moreover, the spherical cubature formula that has been used to develop the CKF has some drawbacks in computation, most notably its inconvenient properties in high-dimensional state estimation problems. To tackle these problems, a new approach to nonlinear state estimation using only an embedded cubature rule, which we have named the square-root embedded cubature Kalman filter (SECKF) is proposed in this work. The experimental results, presented herein, demonstrate the superior performance of the SECKF over conventional nonlinear filters.