Additive divided difference filtering for attitude estimation using modified Rodrigues parameters

In this paper, a real-time attitude estimation algorithm is derived by using an additive divided difference filter as an efficient alternative to the extended Kalman filter. To make the attitude filtering algorithm suitable for real-time applications and to minimize the computational load, a square-root sigma point attitude filter is designed by integrating the divided difference filter with the additive noise concept using the modified Rodrigues attitude parameters. The new attitude filter provides numerically stable and accurate estimates of the state and covariance, but the computational workload of the new attitude estimator is almost identical to the computational complexity of the extended Kalman attitude filter. For performance evaluation the new sigma point attitude filter is compared with the unscented attitude filter and the extended Kalman filter. The sensor measurements include a three-axis magnetometer and rate-gyros. Simulation results indicate that the proposed additive divided difference attitude filter shows faster convergence with accurate and reliable estimation.