Project Details
Projekt Print View

Unambiguity, alternation and non-standard acceptance in automata-based probabilistic model checking

Subject Area Theoretical Computer Science
Term from 2016 to 2024
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 313089026
 
Final Report Year 2024

Final Report Abstract

The project deals with the quantitative analysis of probabilistic systems, modeled by Markov chains and Markov decision processes (MDPs), with respect to linear-time properties represented by temporal logic formulas. The most well-known approach, which has been implemented in several tools, transforms the formula into a deterministic Rabin automaton (DRA) and analyzes the product of the system model and the automaton. The complexity of this approach is double-exponential in the length of the formula and polynomial in the model size. From a complexity-theoretic perspective, this is optimal for MDPs but not for Markov chains. For Markov chains, more efficient algorithms with a single-exponential time bound are known. These include an iterative automaton-free procedure as well as automaton-based approaches with separated (strongly unambiguous) and alternating automata. At the start of the project in 2016, no implementations of the single-exponential methods were known. The main results of the project include analysis algorithms for Markov chains that use unambiguous automata to represent ω-regular specifications. We hereby corrected a flawed algorithm of the literature regarding the use of unambiguous Büchi automata (UBA) for the formal analysis of Markov chains. In addition to the actual UBA-based analysis, algorithms for the generation of UBAs for formulas in LTL (linear temporal logic) and unambiguous automata with other acceptance conditions were also investigated. In comparative experimental studies, the UBA-based approach proved to be competitive with the commonly used approach based on deterministic automaton, and even superior in some cases. Another main goal of the project was to investigate deterministic automata with symbolic representations of Muller acceptance (instead of Rabin acceptance) and their usefulness for the quantitative analysis of both Markov chains and MDPs. A key focus is balancing the advantages that more flexible acceptance conditions can offer in terms of automaton size with the increased computational effort required for graph analysis in the product of the system model and the automaton.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung