Towards a compositional SPIN

  • Authors:
  • Corina S. Păsăreanu;Dimitra Giannakopoulou

  • Affiliations:
  • NASA Ames Research Center, QSS and RIACS, CA;NASA Ames Research Center, QSS and RIACS, CA

  • Venue:
  • SPIN'06 Proceedings of the 13th international conference on Model Checking Software
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper discusses our initial experience with introducing automated assume-guarantee verification based on learning in the SPIN tool. We believe that compositional verification techniques such as assume-guarantee reasoning could complement the state-reduction techniques that SPIN already supports, thus increasing the size of systems that SPIN can handle. We present a “light-weight” approach to evaluating the benefits of learning-based assume-guarantee reasoning in the context of SPIN: we turn our previous implementation of learning into a main program that externally invokes SPIN to provide the model checking-related answers. Despite its performance overheads (which mandate a future implementation within SPIN itself), this approach provides accurate information about the savings in memory. We have experimented with several versions of learning-based assume guarantee reasoning, including a novel heuristic introduced here for generating component assumptions when their environment is unavailable. We illustrate the benefits of learning-based assume-guarantee reasoning in SPIN through the example of a resource arbiter for a spacecraft.