Chiara Troiani - Vision-Aided Inertial Navigation: low computational complexity algorithms with applications to Micro Aerial Vehicles

Accurate egomotion estimation is of utmost importance for any navigation system. Nowadays different sensors are adopted to localize and navigate in unknown environments such as GPS, range sensors, cameras, magnetic field sensors, inertial sensors (IMU). In order to have a robust egomotion estimation, the information of multiple sensors is fused. Although the improvements of technology in providing more accurate sensors, and the efforts of the mobile robotics community in the development of more performant navigation algorithms, there are still open challenges. Furthermore, the growing interest of the robotics community in micro robots and swarm of robots pushes towards the employment of low weight, low cost sensors and low computational complexity algorithms. In this context inertial sensors and monocular cameras, thanks to their complementary characteristics, low weight, low cost and widespread use, represent an interesting sensor suite. This dissertation represents a contribution in the framework of vision-aided inertial navigation and tackles the problems of data association and pose estimation aiming for low computational complexity algorithms applied to MAVs. For what concerns the data association, a novel method to estimate the relative motion between two consecutive camera views is proposed. It only requires the observation of a single feature in the scene and the knowledge of the angular rates from an IMU, under the assumption that the local camera motion lies in a plane perpendicular to the gravity vector. Two very efficient algorithms to remove the outliers of the feature-matching process are provided under the abovementioned motion assumption. In order to generalize the approach to a 6DoF motion, two feature correspondences and gyroscopic data from IMU measurements are necessary. In this case, two algorithms are provided to remove wrong data associations in the feature-matching process. In the case of a monocular camera mounted on a quadrotor vehicle, motion priors from IMU are used to discard wrong estimations. For what concerns the pose estimation problem, this thesis provides a closed form solution which gives the system pose from three natural features observed in a single camera image, once the roll and the pitch angles are obtained by the inertial measurements under the planar ground assumption. Specifically, the system position and attitude can uniquely be determined by observing two point features but improved by exploiting the geometric constraints inherent to a virtual pattern formed by the three features. In order to tackle the pose estimation problem in dark or featureless environments, a system equipped with a monocular camera, inertial sensors and a laser pointer is considered. The system moves in the surrounding of a planar surface and the laser pointer produces a laser spot on the abovementioned surface. The laser spot is observed by the monocular camera and represents the only point feature considered. Through an observability analysis it is demonstrated that the physical quantities which can be determined by exploiting the measurements provided by the aforementioned sensor suite during a short time interval are : the distance of the system from the planar surface ; the component of the system speed that is orthogonal to the planar surface ; the relative orientation of the system with respect to the planar surface ; the orientation of the planar surface with respect to the gravity. A simple recursive method to perform the estimation of all the aforementioned observable quantities is provided. This method is based on a local decomposition of the original system, which separates the observable modes from the rest of the system. All the contributions of this thesis are validated through experimental results using both simulated and real data. Thanks to their low computational complexity, the proposed algorithms are very suitable for real time implementation on systems with limited on-board computation resources. The considered sensor suite is mounted on a quadrotor vehicle but the contributions of this dissertations can be applied to any mobile device.