Symmetric and Efficient Synthesis

Rüdiger Ehlers

Since the formulation of the synthesis problem for reactive systems by Church in the 60s, research on synthesis has lead to both theoretical insights and practical approaches for automatically constructing systems from their specifications. While the first solution of the problem was given by Büchi as early as 1969, only very recently, focus has shifted towards identifying ways to exploit the structure in reactive system specifications in order to lift the scalability of synthesis to industrial-sized designs. The recent progress in synthesis not only lead to a renewed interest in the subject, but also shed light onto the downsides of current synthesis approaches. In the original formulation of the problem, the structure of the produced solutions was not a concern. Experiments with current synthesis approaches has however shown that the computed implementations are usually very hard to understand and have little of the structure that manually constructed implementations have. Furthermore, the scalability of current synthesis approaches is still deemed to be insufficient for many industrial application scenarios, which prevents the introduction of reactive synthesis technology into industrial design flows. In this thesis, we tackle both of these problems for reactive synthesis. To counter the insufficient structure in the solutions, we analyse the problem of symmetric synthesis. In this alternative synthesis problem, the aim is to compute a solution that consists of multiple copies of the same process such that the overall system satisfies the specification. Such systems have no centralised control units, and are considered to be more robust and easier to maintain. We characterise undecidable and decidable cases of the problem, and provide a synthesis algorithm for rotation-symmetric architectures, which capture many cases of practical relevance.

To improve the scalability in synthesis, we start with a simple but scalable approach to reactive synthesis that has shown its principal applicability in the field, and extend its main idea both in terms of scope and usability. We enhance its expressivity in a way that allows to synthesise robust systems, and remove its limitation to specifications of a very special form. Both improvements yield theoretical insights into the synthesis problem: we characterise which specification classes can be supported in synthesis approaches that use parity games with a fixed number of colours as the underlying computation model, and examine the properties of universal very-weak automata, on which we base a synthesis workflow that combines ease of specification with a low complexity of the underlying game solving step. As a side-result, we also obtain the first procedure to translate a formula in linear-time temporal logic (LTL) to a computation tree logic (CTL) formula with only universal path quantifiers, whenever possible.

The new results on symmetric and efficient reactive synthesis are complemented by an easily accessible introductory chapter to the field of reactive synthesis that can also be read in isolation.

PhD Thesis.