Unscented Kalman filtering for relative attitude and position estimation
Goh, Shu Ting
MetadataShow full item record
In this thesis, the Unscented Filter is derived for the relative attitude and position estimation. In the previous work, the Extended Kalman Filter had shown its success in the estimation. For an inaccurate nonlinear model system, the Unscented Filter has proved to converge faster than the standard Extended Kalman Filter. The relative attitude and position estimation approach is based on the vision-based navigation (VISNAV) system. The line-of-sight measurement with the gyro measurements are used by the Unscented Filter to estimate the relative attitude, gyro biases and relative position. The modified Rodrigues parameters are used to define the local quaternion error for the sigma points transformation, while the quaternion is used for global attitude propagation. The transformation between quaternion and modified Rodrigues parameters is always guarantees quaternion normalization. Both spacecraft are assumed to have an equal type of build so that their orbits are predictable. The simulation results show that the standard Extended Kalman Filter is only able to provide converged results in a very limited low error, while the Unscented Filter can provide converged results up to a certain high error and with a faster convergence rate.