|Rohloff, Tom: Flow control for event-based cyber-physical systems. |
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Masterarbeit Nr. 72 (2018).
75 Seiten, englisch.
Cyber-physical systems (CPS) gained widespread attention in recent years. In the context of Industry 4.0, they are often implemented as Networked Control Systems (NCS) that use a shared network. Past research already investigated stability, as well standard performance criteria of the control system. Based on different trigger functions of previous work, the question is addressed how their trigger conditions should be parameterized to obtain optimal results with respect to the performance. This thesis focuses on discrete-time Linear Quadratic Regulators (LQR) that operate in bandwidth-limited networks. Leveraging the quadratic nature of the problem, the optimization technique proposed uses gradient descent and heuristics based on other metrics gathered from the network. To evaluate this approach, a simulation framework is developed. The quantitative analysis shows that a proposed trigger parameter adoption policy performs very well. The overall control system performance is only marginally worse than the optimum (about 4 % on average). Compared to the naive approach of triggering at every sampling instant, performance gains ranging from 6 % to over 500 % (depending on the bandwidth limit) could be achieved. Additionally, the control system performance stays consistent, even in the presence of extreme constraints of the relative link capacity.
|Abteilung(en)||Universität Stuttgart, Institut für Parallele und Verteilte Systeme, Verteilte Systeme|
|Betreuer||Rothermel, Prof. Kurt; Carabelli, Ben|
|Eingabedatum||5. Juni 2019|