A Framework for Integrity Monitoring for Positioning through Graph-based SLAM Optimization

As satellite navigation show vulnerabilities in specific circumstances such as urban canyons or jamming and spoofing situations, additional sensors such as cameras may be incorporated on the platform. Despite advancements in the robotics and computer vision community, which have lead to increasingly accurate Simultaneous Localisation and Mapping (SLAM) positioning solutions, visual navigation has its own vulnerabilities. It remains therefore of critical importance for many applications to study the integrity of fused navigation algorithms and their components, which is done less for SLAM than for satellite navigation. In this paper, a framework for integrity monitoring (IM) of a visual SLAM algorithm is proposed. A sensor-level IM scheme analyses feature reprojection errors. It is demonstrated that in dynamic environments multiple hypotheses can be generated from different subsets of feature measurements. Additionally, the factor graph-based framework employs a fusion-level IM scheme which deals with these multiple hypotheses and selects the most probable one by calculating the sum of weighted measurement residuals. These concepts are applied to scenarios from real and simulated experiments in order to demonstrate applicability.