Method of attitude estimation in magnetometer and accelerometer system based on MRPs and NPUPF

For the problem of existing attitude estimation algorithms based on geomagnetic and accelerometer, what there are shortages on covariance matrix´s singularity and exactly known local geomagnetic vectors, a new basic attitude estimation algorithm is proposed. This new algorithm uses Modified Rodrigues Parameters to express the dynamics of system, so that the singularity of covariance matrix owing to quaternion´s redundancy in expression of system´s dynamics is eliminated, and the geomagnetic vector as a stationary process is put into the estate variables because of its slowly varying character, so that the problem of exactly known local geomagnetic vectors is solved. In allusion to the performance degradation of attitude estimation method based on MRP and Unscented Particle Filter when there are large model error and colored noise, an improved scheme for attitude estimation based on Nonlinear Predictive Unscented Particle Filter is proposed. Simulation results verify that the method improves the robustness and environmental adaptability of attitude estimation algorithm under the condition of system with large model error and colored noise.