A Robust Visual-Aided Inertial Navigation Algorithm for Pedestrians

Figure 2

Algorithm framework illustrating the full pipelines. Odometry tracking outputs the odometry navigation results and looping information. In ZUPT-aided INS, zero-velocity measurements are fused with SINS in an Extended Kalman Filter to reduce error growth. Pose graph optimization continuously optimizes the pose graph based on both VO/VIO pose estimate and SINS position output. Parameter adaptive adjustment module first analyses the error between truth values and navigation results and construct optimization problem to solve accurate zero-velocity detection thresholds.