Real- Time State Estimator with Multi-Sensor Fusion Based on Point-and- Line Construction

As robots are required to be more robust and practical, relying solely on a single sensor suite is no longer sufficient. Therefore, it is crucial and practical to adaptively switch between different sensors depending on the complexity of the scenes, in order to effectively satisfy user needs. To address these challenges, we introduce a versatile real-time state estimator that relies on point-line features. The salient contributions of this research can be condensed into the following points: 1) the algorithm incorporates a pyramid selection mechanism to reduce feature detection time while supporting the integration of monocular-inertial, stereo-inertial, and stereo camera systems; 2) the error objective function is constructed by adding line segment detector and line segment residuals to the multi-sensor selection frame; 3) the experimental outcomes demonstrate that our novel system efficiently diminishes the duration for extracting line features from the EuRoC dataset. When compared to the state-of-the-art methods, the proposed algorithm demonstrates superiority in enhancing the precision of map localization as well as broadening the scope of scene applicability.