Modular C++ Library for Relaxed Unscented Kalman-Filtering

Abstract

Filtering and sensor fusion are critical tasks in advanced engineering applica- tions, especially in robotics and autonomous vehicles. It is a general problem that the high accuracy and low computational cost are mutually exclusive in such filtering algorithms. The Unscented Kalman-Filter (UKF) is a golden mean between the Extended Kalman- Filter (EKF) and the Particle Filter. Recently, the authors have proposed a generic com- putational relaxation for the EKF that provides options to decrease the computational cost by exploiting the partially linear nature of the mappings in the system model. This pa- per introduces an open-source C++ RelaxedUnscentedTransformation library that fully implements the proposed method. Since the technique offers several independent usage options, different components are implemented, and the corresponding use cases are il- lustrated through examples. Via numerical tests, the paper shows that the implementation can significantly decrease computational costs and even provide an opportunity to in- crease filtering accuracy.