Vision-Aided Inertial Navigation With Linear Feature Constraints

The patent segments the computational task by processing visual features independently and forming constraints for each feature separately. Each visual feature generates its own pose constraint equation, which can be processed and accumulated independently. This segmentation allows the system to handle multiple features without exponentially increasing computational complexity, as each feature contributes additively to the overall constraint set rather than creating combinatorial interactions.

The patent applies dynamics by treating the detection status of visual features as a time-varying property. The system continuously monitors whether detected features are static or moving, and dynamically adjusts which features are used for navigation constraints. Features identified as moving (potential obstacles) are excluded from constraint formation, while static features continue to provide localization information. This dynamic classification allows the system to maintain navigation accuracy using static features while simultaneously detecting moving obstacles that are excluded from the constraint set.