Decreasing the Computational Demand of Unscented Kalman Filter based Methods

Abstract

Computational load is a critical factor in sensor fusion applications especially in mobile devices (e.g., robots, drones, etc.) with limited resources onboard. The paper proposes a computational relaxation for the Unscented Transformation (UT) that is an essential part of the Unscented Kalman-filter based applications. The derivation for the most commonly used UT variant is presented and it is shown how the number of necessary sigma points is reduced. The practical merit of the proposed relaxation is demonstrated through a mobile robot localization example that clearly shows the benefit in terms of CPU time.