Odometry and trajectory estimation approach for unobserved and autonomous vehicles

All different kinds of unobserved vehicles are omnipresent in modern society. Autonomous driving in traffic is intensely researched, and Unmanned Aerial Vehicles (UAVs) for logistical or even military purposes are more present than ever before. Even in motorsport, there are autonomous racing events nowadays. All of those require the position of a moving object to be determined. Even decades ago, tracking trajectories was already an important part of space exploration.
This thesis aims to investigate how odometry and trajectory estimation can be simulated and implemented with minimum sensor effort for miniaturized vehicles moving in two dimensions.
To achieve this, so-called Inertial Measurement Units (IMUs), which are a fusion of sensors such as accelerometers, gyroscopes, and magnetometers, are used. Any additional information about the position, which may be determined by GPS, RADAR, LIDAR or mobile communications in state-of-the-art technologies, is omitted. The usefulness, reliability and accuracy of position estimation in the absence of this information is investigated.
In addition to simulating and estimating ideal trajectories, the thesis also covers aspects of Video-Object-Tracking (VOT) and the evaluation of real measurements recorded with a prototype vehicle. With simulations, estimation errors of less than 1 % of the trajectories length and less than 2 cm accuracy are achieved, whereas the errors for real measurements are strongly influenced by non-ideal effects investigated in this thesis, as well as the trajectory shapes themselves, resulting in errors easily exceeding 15 %