Doctoral College Cyber-Physical Production Systems at TU Wien
> Zum Inhalt


Production systems are extensively equipped with sensors, actuators and controllers, and the automation pyramid is linked through a communication network into an either strictly centralized or to a strictly decentralized IT structure. A fixed set of products and services is typically offered, and their realization is carefully planned in advance: Unfortunately, too sequential, too comprehensive, and hardware/product specific. Akin to a biological system, a CPPS will adapt itself in real time to its environment, reshape on the fly its products and production lines, and lower its costs and pollution footprint. While many advances are already under way, there are still many research and technological challenges that have to be mastered. One of the greatest challenges is to efficiently predict the emergent behaviors of these systems. However, the complexity of their models often hinders any attempt to exhaustively verify their safe behavior.

An alternative method is to equip them with smart controller to predict emergent behaviors at runtime. This approach makes Cyber-Physical Systems self-aware opening up new perspectives to design smart systems able to dynamically self-organize or reconfigure themselves in order to adapt to the different circumstances. 

Hence, in this project, our work can be divided into following parts:

Model of System: We are going to use data (input/output and execution traces) in order to learn Dynamic-Bayesian-Network (DBN) models for the CPS systems. A particular form of DBNs that we are going to consider are Neural-Circuit DBNs (NC-DBN), that is DBNs whose nodes behavior is represented by neurons, as they are found in the neural circuits of the C.elegans worm. An abstract example of this kind of neural circuit is depicted in Figure 1.

Figure 1: Simple neural circuit instance

Usage of the Model: Estimating the internal state of the system Predict System near-future actions.

Goals of this Project: Building Dependable Systems, Robust System and Smart Controllers for CPPS. These controllers will encode all the valid behaviors within one stochastic program, whose adaptation to environment perturbations will merely prune away some of the previously valid behaviors.

Experiments: We plan to implement the parallel parking on mobile robot, which is a kind of cyber-physical systems and depicted in Figure 2.

Figure 2: Mobile robot - pioneer rover

To fully validate our methods, we may also implement algorithms on other platform based on our further investigations.

PhD-Student and Supervision

PhD-Student: Guodong Wang, M. Sc., B. Sc.

Guodong Wang comes from a small and beautiful county, named Shangcheng, China. From 09/2011 to 07/2014, he studied at Nanjing University of Information Science & Technology, China, as a Master student in the Master program "Systems Analysis and Integration", which is mainly focused on Complex Systems Analysis and Modeling. His research topic was "Self-Adaptivity of Cyber-Physical Systems". During his studies, he worked as a research assistant and participated in many projects, e.g., a project on "Intelligent Vehicles". From 09/2007 to 07/2011, Guodong Wang studied at Chenggong College, Henan University of Economics and Law as a bachelor Student with a major in "Computer Science and Technology". During the four years, Guodong’s major interests were Software Engineering and Programming Languages. Since 03/2015, Guodong is working on his PhD as a research assistant at the Cyber-Physical Systems Group, Institute of Computer Engineering, Vienna University of Technology, supervised by Prof. Radu Grosu and co-supervised by Assistant Professor Ezio Bartocci. Currently, his research interests contain: Deep Neural Network, Neuroscience, Distributed Multi-Agent Computing, CPS.


Advisor: Prof. Dr. Radu Grosu, Institute of Computer Engineering (E182)

Co-Supervisor: Dr. Ezio Bartocci, Institute of Computer Engineering (E182)