Inertial aided monocular SLAM for GPS-denied navigation

This paper addresses the problem of navigation in GPS-denied scenarios using a low-cost inertial measurement unit (IMU) and monocular video data. It is shown how the standard error state Kalman-Filter approach for GPS-INS integration can be extended to take advantage of imagery measurements by augmenting the filter state with the coordinates of landmarks in inverse depth parameterization. This yields a robust monocular SLAM algorithm that computes the desired navigation solution and a sparse feature map at the same time and is suitable for real-time operation. Results are presented which demonstrate the applicability of this approach for indoor and outdoor scenarios.