Minimal Assumptions Refinement for Realizable Specifications

Abstract

A challenge that has gathered much attention in recent years is automated synthesis of correct-by-construction software systems from declarative specifications. The specification language is typically a subset of linear temporal logic called generalized reactivity of rank 1, for which there exists an efficient synthesis algorithm. Specifications in this language model the system as the interaction between an environment and a controller, the former satisfying a set of assumptions and the latter a set of guarantees. In order for a solution to exist, a sufficient set of assumptions implying the guarantees must be provided. The assumptions must be as general as possible and small enough to be intelligible by engineers that need to assess their consistency with the true environment where the synthesized controller will operate. The search for such assumptions is generally a refinement approach driven by counterstrategies, characterizations of undesirable environment behaviors that force the violation of the guarantees; assumptions are progressively refined in order to exclude such behaviors. In this work we provide a heuristic to drive this counterstrategy-guided search towards smaller refinements. We define a concept of minimality of refinements with respect to counterstrategies and provide an algorithm that provably finds minimal refinements with little time overhead. We show experimentally that it consistently produces one or more shorter solutions than state of the art for a set of popular case studies. We also demonstrate that in a popular case study (AMBA-AHB protocol) our heuristic finds a close-to-optimal solution that cannot be found by previous fully automated approaches. CCS CONCEPTS • Software and its engineering Formalsoftwareverification; Requirements analysis; • Hardware Buses and high-speed links.

Combining Genetic Programming and Model Checking to Generate Environment Assumptions

“…Learning Assumptions. The problem of automatically inferring assumptions of software components, a.k.a supervisory control problem, was widely studied in the literature (e.g., [1], [7]- [9], [67], [80]- [84]). However, the solutions proposed in the literature are solely focused on components specified in finite-state machines and are not applicable to signal-based formalisms (e.g., Simulink ® models), that are widely used to specify CPS components (see Section I).…”

“…The problem of synthesizing environment assumptions has been extensively studied in the area of formal verification and compositional reasoning (e.g., [3]- [7]). There have been approaches to automate the generation of environment assumptions in the context of assume-guarantee reasoning using an exact learning algorithm for regular languages and finite state automata [1], [8], [9].…”

Spectra: a specification language for reactive systems

“…In the first two project classes, in 2015 and in 2017, the students have developed about 10 different autonomous Lego robots, which they wrote the specifications for, and actually built and run. 14 In the third and fourth project classes, in 2019 and 2020, the students have developed several example systems with a PC-based simulation. From these classes, we have collected over 320 versions of specifications, all written by these students as they worked on their projects for several months.…”

“…Indeed, we have already used SYNTECH15 and SYNTECH17 as benchmarks to examine the motivation for and the efficiency and effectiveness of various analyses, e.g., [28,43,51,53]. Others have also already used them, e.g., [14,59]. As we make these collections publicly available, we hope they will serve to advance reactive synthesis research and its future use in practice.…”

Unrealizable Cores for Reactive Systems Specifications

“…Other works have considered means to repair unrealizable specifications by automatically suggesting additional assumptions that will make the specification realizable, see, e.g., [2], [8], [9], [10], [21]. It may be possible to combine the computation of a core or of all cores with a repair approach.…”