A System for Evaluating Vision-aided Navigation Uncertainty

To enable the fusion of multiple different navigation solutions, accurate uncertainty information about each input navigation solution must be known. Therefore, it is beneficial to study the uncertainty of each possible navigation solution and ensure it generates predicted uncertainty estimates that comport with reality. In the past, studies of uncertainty in vision-aided navigation have primarily been conducted in simulation. In this paper, we present a system that can enable researchers to study the uncertainty of vision-aided navigation systems by generating a large quantity of data with associated “truth”.