Parallel Quantitative Analysis of Probabilistic Systems

Bc. Milan Češka, Ph.D.

Master's thesis

Abstract:

This master thesis deals with an automatic technique for verification of finite probabilistic systems based on an explicit model checking. The thesis presents a problem of parallel quantitative analysis of probabilistic models against a specification in linear temporal logic. Semantics of such a model is given by a finite Markov decision process. The given problem is transformed into a graph problem and the consequent construction and solution of the corresponding linear programming problem, called the liner program. The size and structure of the linear program has a significant effect on the time complexity of the designed solution. New methods for minimization and parallel processing of the linear program are proposed. The main contribution of this thesis is design and implementation of a tool for parallel quantitative analysis of probabilistic systems running in a shared memory environment. The tool builds on the parallel algorithm for a reachability analysis and a graph decomposition into strongly connected components. Several experiments were done to verify whether the propose methods and their implementation reduce the run time of the designed tool.

Keywords

automatic verification, parallel model checking, linear temporal logic, probabilistic models, Markov decision processes, qualitative analysis, strongly connected components, end components, system of linear inequalities, linear programming

Language used: English

Date on which the thesis was submitted / produced: 23. 5. 2008

Thesis defence

• Date of defence: 3. 7. 2008
• Supervisor: prof. RNDr. Jiří Barnat, Ph.D.
• Reader: prof. RNDr. Ivana Černá, CSc.

Masaryk University

Faculty of Informatics

Master programme / field:
Informatics / Informatics