Sequential attitude estimation using particle filters

An efficient attitude estimation approach is derived by utilizing particle filtering. The new attitude particle filtering algorithm is formulated in terms of the quaternion parameters and maintains the unit-norm constraint naturally without modification. This work investigates a number of improvements on particle filters that are developed independently in various engineering fields. Several variants of the particle filter include the regularized particle filter, unscented particle filter, and Markov chain Monte Carlo particle filter. The performance of the quaternion particle filter is compared with that of the extended Kalman filter and recently proposed unscented Kalman filter through a simulation case involving a low earth-orbiting spacecraft acquiring measurements from the magnetometer and rate-gyro sensors. The simulation results indicate that the flexible nature of the particle filtering renders the quaternion-based particle filter being more adaptive to some features of the complex attitude systems, leading to faster convergence.