English

Fully Generalized Reactivity(1) Synthesis

Formal Languages and Automata Theory 2024-02-06 v1

Abstract

Generalized Reactivity(1) (GR(1)) synthesis is a reactive synthesis approach in which the specification is split into two parts: a symbolic game graph, describing the safe transitions of a system, a liveness specification in a subset of Linear Temporal Logic (LTL) on top of it. Many specifications can naturally be written in this restricted form, and the restriction gives rise to a scalable synthesis procedure -- the reasons for the high popularity of the approach. For specifications even slightly beyond GR(1), however, the approach is inapplicable. This necessitates a transition to synthesizers for full LTL specifications, introducing a huge efficiency drop. This paper proposes a synthesis approach that smoothly bridges the efficiency gap from GR(1) to LTL by unifying synthesis for both classes of specifications. The approach leverages a recently introduced canonical representation of omega-regular languages based on a chain of good-for-games co-B\"uchi automata (COCOA). By constructing COCOA for the liveness part of a specification, we can then build a fixpoint formula that can be efficiently evaluated on the symbolic game graph. The COCOA-based synthesis approach outperforms standard approaches and retains the efficiency of GR(1) synthesis for specifications in GR(1) form and those with few non-GR(1) specification parts.

Keywords

Cite

@article{arxiv.2402.02979,
  title  = {Fully Generalized Reactivity(1) Synthesis},
  author = {Rüdiger Ehlers and Ayrat Khalimov},
  journal= {arXiv preprint arXiv:2402.02979},
  year   = {2024}
}

Comments

This is an extended version of the paper accepted at TACAS'24

R2 v1 2026-06-28T14:38:29.564Z