The main target of this project is to build an autonomous boat which can navigate through a simultaneously created map or a given one. The development of this boat was made possible through a cooperation with the Miniatur Wunderland Hamburg. A …
Machine learning is more and more applied in our everyday life, but also safety critical systems are increasingly equipped with ML procedures. This paper gives an insight into the implementation of a machine learning co-processor for embedded systems …
In the research project FAUST at the University of Applied Sciences Hamburg driver assistance and autonomous systems are developed. The control algorithms of these systems have the ability to control them under certain conditions. If the conditions …
This work shows a way to implement an agent, capable to learn to drive a given track with a high speed. To illustrate the need of a machine learning technique, the physics and the complexity of driving a turn are shown. After this there is a …
In this work a model-free path tracking method was developed. The method considers the vehicle kinematics, the path geometry and the delayed control actions in order to determine the optimal control action. The predictive path tracking method applies …
Im Forschungsprojekt FAUST aus dem Department für Informatik der Hochschule für Angewandte Wissenschaften Hamburg werden Technologien für Fahrerassistenz- und Au- tonome Systeme entwickelt und entworfen (FAUST, 2010). Autonomes Fahren wurde auf …
We propose local error estimates together with algorithms for adap- tive a-posteriori grid and time refinement in reinforcement learn- ing. We consider a deterministic system with continuous state and time with infinite horizon discounted cost …
We develop a theoretical framework for the problem of learning optimal control. We consider a discounted infinite horizon deterministic control problem in the reinforcement learning context. The main objective is to approximate the optimal value …
Reinforcement learning methods for discrete and semi-Markov de- cision problems such as Real-Time Dynamic Programming can be generalized for Controlled Diffusion Processes. The optimal control problem reduces to a boundary value problem for a fully …