The Unscented Kalman Filter With Reduced Computation Time for Estimating the Attitude of the Attitude and Heading Reference System

Abstract

The algorithms of the Kalman filters have been used in many papers on the Pedestrian Dead Reckoning (PDR) and attitude estimation for the attitude and heading reference system (AHRS). In this article, one type of the nonlinear Kalman filters, the Unscented Kalman filter (UKF) was researched to reduce computational cost, while maintaining accuracy. One of the issues of the attitude estimation algorithms is that computational cost is large, because of many matrix computations. The computational cost should be reduced for the application of the navigation system for general consumers toward developing low-priced navigation system. In this article, the novel UKF, named “Kaisoku Unscented Kalman Filter (KUKF)” is proposed. It was verified that the proposed KUKF reduced the computational cost about 13.426% comparing with the existing UKF, while almost maintaining accuracy.