Aided Visual Odometry: A Sparse Bundle Adjustment Solution for Rover Navigation

Acquisition of accurate ego-motion estimates sits at the core of autonomous navigational re- search. In this thesis, a method of estimating rover trajectory through stereo visual odometry is proposed. This research presents a navigational system aided by a sun-sensor and inclinome- ter to improve accuracy in motion estimation. Specifically, this research improves on the work conducted by Lambert et al. [1] and investigates the accuracy of previous work as well as the quality of the Devon Island dataset. Sparse implementations of matrices are utilized to effi- ciently determine trajectory estimations using both stereo camera and the secondary attitude sensors. Ultimately this thesis implements an algorithm that results in a significant reduction in computational cost when compared to the work conducted by Lambert et al. [1] iii