Control System Theoretic Approach to Model Based Navigation

This thesis gives an overview and classification of current conventional and model based navigation algorithms found in the literature. Model based in that context means, that a mathematical/physical system model of the moving vehicle is integrated into the Kalman Filter, which normally performs the state estimation task. In contrast to conventional navigation algorithms, which rely solely on inertial measurements for the prediction of states, model based approaches have the great advantage, that a prediction is possible even when no measurements are available. The architecture of the new Full State Vehicle Dynamics (FSVD) algorithm is based on the principals of state estimation using a Luenberger observer or Kalman Filter, which in general runs parallel to the technical system, whose states need to be estimated. Current conventional and model based navigation algorithms are described and compared mathematically.
In a next step a comprehensive evaluation of all algorithms took place. Therefore two different examples are given. On the one hand an autonomous ying vehicle which is evaluated in simulations and on the other hand a real navigation test vehicle, which is a van with numerous sensors on board. Low cost sensors which aid the model based navigation solution and high-end sensors which serve as reference system for validation. For both systems the mathematical system description is given and results are presented.