A new method for generating sigma points and weights for nonlinear filtering

In this paper, a new method termed as new sigma point Kalman filter (NSKF), is proposed for generating sigma points and weights for estimating the states of a stochastic nonlinear dynamic system. The sigma points and their corresponding weights are generated such that the points nearer to the mean (in inner product sense) have a higher probability of occurrence, and the mean vector and covariance matrix are matched exactly. Performance of the new algorithm is compared with the existing unscented Kalman filter (UKF), the cubature Kalman filter (CKF), the cubature quadrature Kalman filter (CQKF) and higher order unscented filter (HOUF) for two different problems. Comparison is done by calculating the root mean square error (RMSE), relative computational time and track-loss. From simulation results, it can be concluded that the proposed algorithm performs with superior estimation accuracy when compared to the UKF, CKF, CQKF and HOUF