A Nonlinear Bingham Filter for Pose Using Dual-Quaternions

Open Access
- Author:
- Goodyear, Andrew
- Graduate Program:
- Aerospace Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- October 06, 2022
- Committee Members:
- David Bradley Spencer, Dissertation Advisor
Robert G. Melton, Committee Member
Puneet Singla, Chair of Committee
Joseph Paul Cusumano, Outside Member
Amy Ruth Pritchett, Program Head/Chair - Keywords:
- Estimation
Filtering
Dual-Quaternion
Relative Motion
Navigation
Pose - Abstract:
- Simultaneous estimation of both position and orientation, commonly referred to in the literature as pose, is a subject of continuing interest in the aerospace research community and industry. With the advent of large, proliferated constellations in low earth orbit, the ability for a satellite to track and estimate relative pose in autonomous proximity operations is desired. A way to parameterize pose in a compact and singularity-free way that is computationally efficient has given rise to study into an older way to represent kinematics and rigid body motion: dual-quaternions. Dual-quaternions have found their niche in computer graphics applications due to their numerical efficiency, but there are issues involved with using them for real-time pose estimation. The issue with dual- quaternions is that they do not exist within a standard real number space, meaning the usage of more exotic distributions to describe dual-quaternion uncertainty must be used. The traditional solution to this issue is to make a small-angle assumption, meaning an estimation filter will likely break down if the estimate strays too far from the truth. This dissertation addresses these issues with dual-quaternion filtering using two approaches. The first approach is a constrained extended Kalman filter (EKF) and the second is a Bingham-distributed unscented Kalman filter (UKF). The development of both these filters is provided, and they are demonstrated to accurately spacecraft relative pose through simulation.