Visual-Inertial Tightly Coupled Fusion and Nonlinear Optimization for UAVs Navigation

Visual-Inertial fusing for UAVs navigation has been an attractive complementary solution with great potential. Visual approaches can provide lots of information about the environments but will fail in less-features surroundings or in fast motion. What’s more, visual sensor has no sense of gravity vector and monocular has no observability of metric scales. IMU can predict the pose of camera accurately and supply gravity vector for visual-ways localization. In this literature, we propose a novel visual-inertial tightly fusion and nonlinear optimization approach for UAVs navigation. Camera-IMU calibration is carried on for real experiment and camera-IMU states initialization is also considered as optimization can benefit from a good initial states guess. Experiments are implemented to evaluate the algorithms using EuRoC flying datasets and our own designed flying platform. Performances are compared with the state-of-the-art visual SLAM approach.