Synthesis of Asynchronous Distributed Systems from Global Specifications
Jesko Hecking-Harbusch
The synthesis problem asks whether there exists an implementation for a given formal specification and derives such an implementation if it exists. This approach enables engineers to think on a more abstract level about what a system should achieve instead of how it should accomplish its goal. The synthesis problem is often represented by a game between system players and environment players. Petri games define the synthesis problem for asynchronous distributed systems with causal memory. So far, decidability results for Petri games are mainly obtained for local winning conditions, which is limiting as global properties like mutual exclusion cannot be expressed.
In this thesis, we make two contributions. First, we present decidability and undecidability results for Petri games with global winning conditions. The global safety winning condition of bad markings defines markings that the players have to avoid. We prove that the existence of a winning strategy for the system players in Petri games with a bounded number of system players, at most one environment player, and bad markings is decidable. The global liveness winning condition of good markings defines markings that the players have to reach. We prove that the existence of a winning strategy for the system players in Petri games with at least two system players, at least three environment players, and good markings is undecidable.
Second, we present semi-decision procedures to find winning strategies for the system players in Petri games with global winning conditions and without restrictions on the distribution of players. The distributed nature of Petri games is employed by proposing encodings with true concurrency. We implement the semi-decision procedures in a corresponding tool.
PhD Thesis.