A New Robust Square-Root Unscented Kalman Filter

Abstract

As an improved unscented kalman filter, square-root unscented kalman filter has better robustness than unscented kalman filter, but there is still collapse phenomenon in existence when square-root unscented kalman filter is applied. In order to eliminate filtering collapse, a new robust square-root unscented kalman filter is presented. This new filter adopts more appropriate filtering algorithm structure, avoids calculating covariance matrix frequently to reduces the computational complexity, uses Cholesky decomposition rank one update to calculate the square root of covariance matrix. Then a simulation is hold to validate the performance of new filter. The simulation result demonstrated that the new robust square-root unscented kalman filter whose filtering precision is equal to unscented kalman filter has better robustness.