Datenbestand vom 13. Juni 2019
Tel: 089 / 66060798
Mo - Fr, 9 - 12 Uhr
Fax: 089 / 66060799
DER VERLAG IST IN DER ZEIT VOM 12.06.2019 BIS 23.06.2019 AUSCHLIESSLICH PER EMAIL ERREICHBAR.
aktualisiert am 13. Juni 2019
978-3-8439-3652-1, Reihe Informatik
Modeling Robotic Systems with Activity Flow Graphs
152 Seiten, Dissertation Eberhard-Karls-Universität Tübingen (2017), Softcover, A5
Intelligent systems have to perceive their surroundings in order to facilitate autonomy. With an understanding of the environment, they can make their own decisions based on high level control policies defined by the developers. Robotic systems differ drastically in their sensory capabilities, their computational power, and their designated tasks. When developing algorithms, however, we need to have a common modeling framework that enables us to generalize and re-use existing solutions. A modular approach, which is coherent across different platforms, also allows faster prototyping of new systems. In this dissertation we develop a modeling framework based on data flow that achieves this goal.
We first extend the existing Synchronous Data Flow (SDF) model and combine it with reactive programming and finite-state machines. Together, these existing frameworks enable us to model many aspects of complex robotic systems. We apply this model to a robot in a warehouse scenario to demonstrate the viability of the approach. We then merge SDF and reactive programming into Hybrid Flow Graphs (HFGs), where we explicitly model synchronous and asynchronous data flow. We then apply the HFG model to the perception system of an autonomous transportation robot.
In a last step, we eliminate the need for separate finite-state machines by introducing the concept of activity into the data flow. We unify the different aspects into a coherent framework which we call Activity Flow Graphs (AFGs). The result is a single computation graph that can express both perception and high level control aspects of any robotic system. We then demonstrate this with multiple high level robotic system models.
Finally, we make use of the uniform AFG model to provide a graphical user interface that allows a developer to rapidly prototype complete robotic systems. Since all aspects of a robot can be implemented using the same framework, there is no need to switch between different paradigms. The user interface is designed to give immediate feedback, which speeds up prototyping, testing and evaluation, as well as debugging when working with real robots.