Advanced Filter Design for Autonomous Vehicles

Abstract: The introduction of autonomous vehicles comes with many benefits related to safety and quality-of-life, the implementation of which is a challenging task for engineers to solve; one part of this aforementioned task is the requirement for precise positioning.Firstly, this thesis work investigated the contemporary fields of science rel-evant for the task of position estimation. The paper then dissertated the theory, implementation and the experimental evaluation of three di˙erent methods of estimating position. The individual properties was then examined for the dif-ferent vehicle filters and the optimal method of estimating position was then determined with regards to its specified use case, the autonomous truck appli-cation.The methods chosen for evaluation were the Unscented Kalman Filter (UKF), the Controller Output Observer (COO) and the Washout filter method. These methods were evaluated using physical experiments carried out on roads and the result of which showed that the Controller Output Observer (COO) and the Washout filter methods shared significant disadvantages compared to the supe-rior Unscented Kalman Filter (UKF). Based on the experimental results a new filter constellation were proposed whereby two modified position evaluating filters are connected in series.