Theses

I offer supervision for Bachelor and Master theses. If you are interested in a bachelor or master thesis (on simulation or on any other topic), you are welcome to contact me.

Topics for Theses

Possible topics for theses:

Simulation-based optimization of network protocols (Bachelor)

  • Status quo: there exist various network protocols with different properties for different areas of application.
  • Objective of the thesis: development of a simulation environment to test and optimize various network protocols, investigation of performance in terms of latency, throughput and packet loss
  • Open questions: Which layer(s) of the OSI model are interesting or relevant? Which protocols should be examined? Which parameters can be measured? How can the simulation be made realistic? How can the protocols be adapted to achieve better performance?

Comparison of time-based and event-based simulation based on case studies (Bachelor)

  • Status quo: there are several formally defined case study models for stochastic hybrid systems with different properties from the ARCH Friendly Competition, which have been simulated and/or analyzed with different tools so far.
  • Objective of the thesis: development of time-based and event-based simulations for ARCH case studies, comparison of the speed and accuracy of the simulations
  • Open questions: Which case studies are suitable for time-based and event-based simulation? What are the limits of practical feasibility? Which parameters influence performance?

Development of a statistical model checker for rectangular autoata (Master)

  • Status quo: First approaches for statistical model checking of singular automata exist; rectangular automata additionally allow continuous non-determinism in the form of intervals in invariants, derivatives and jump conditions. This makes the analysis of rectangular automata difficult and inefficient for models with many random variables.
  • Objective of the thesis: Development of an event-based simulation approach for rectangular automata, including dealing with non-determinism (by resolution or by computing optimal schedulers) and the extension of the formalism with stochastic variables / random clocks, and addition of statistical model checking methods
  • Open questions: What are the challenges of non-determinism in rectangular automata?  How can we deal with non-determinism in simulation? How can previous simulation approaches be transferred? How easily can the formalism be extended to include stochastic variables, e.g. random clocks? How can statistical model checking methods be applied? How easy would an extension to linear hybrid automata (according to Henzinger) be?