Absolute and Relative Estimation Techniques in an Aerial-Ground Multi-Robot System

Absolute orientation estimation is a key component in many fields such as aerospace, robotics, navigation and rehabilitation. A common solution to such problem consists of employing a 9-axis IMU (Inertial Measurement Unit) composed of a 3-axis accelerometer, a 3-axis gyroscope and a 3-axis magnetometer in order to estimate the RPY (Roll-Pitch-Yaw) angles of a rigid body by first integrating the gyroscope measurements, and then by correcting the drifts present in the previous estimates because of the gyroscope biases through the measurements of the gravity vector direction and of the earth's magnetic field direction. Although simple, superficial implementation of this approach may lead to inaccurate estimates as the gravity vector measurement can be influenced by unexpected linear acceleration, and the earth's magnetic field measurement can be perturbed by external magnetic fields such as those generated by ferromagnetic materials. The first main topic of this thesis consists in the study of a technique able to provide a robust solution to the absolute orientation estimation problem against the previous issues. In particular, it is analysed a quaternion-based sensor fusion algorithm having in mind as the target application the absolute orientation estimation of a UAV (Unmanned Aerial Vehicle). Autonomous landing on mobile platforms represents a significant improvement in the drones use. Several researches have been conducted in order to study different solutions for implementing such manoeuvre. Many of these often rely on the use of a GNSS (Global Navigation Satellite System) module, however this kind of approach not always is possible as the GNSS information in some cases may be partially deteriorated and in others it may be completely unavailable. A common method in these situations is to equip the UAV with an onboard camera and to rely to vision information. The second main topic of this thesis consists in the estimation of the relative position and velocity of a moving UGV (Unmanned Ground Vehicle) with respect to a UAV in a GNSS-denied environment in order to allow the UAV to land on the UGV. The problem is tackled by mainly exploiting the relative bearing of the UGV with respect to the UAV, which may be practically considered as an information provided by a camera.