High Order Nonlinear Least-Squares for Satellite Pose Estimation

This paper introduces a high-order nonlinear least-squares method for solving six-degree-of-freedom (6-DOF) navigation of satellite maneuvers. The approach involves developing first through fourth-order Taylor series models, which provide the necessary conditions that are iteratively adjusted to recover the unknown roots for reducing the errors arising from fitting models to a given set of observations. An initial guess is provided for the unknown parameters in the system, developing a correction vector using Taylor expansion, and then manipulating the necessary conditions to provide the least-squares algorithm. Computational differentiation (CD) tools generate the Taylor expansion partial derivative models. This paper presents an experimental work conducted using a 6-DOF platform to demonstrate the performance of the developed high-order nonlinear least-squares navigation method. An initial calibration of the sensing systems is performed in an operationally relevant ground-based environment. The experiments demonstrate that accelerated convergence is achieved for the second-, third-, and fourth-order expansions with various initial guess conditions.