Using Fairness to Make Abstractions Work

Abstractions often introduce infinite traces which have no corresponding traces at the concrete level and can lead to the failure of the verification. Refinement does not always help to eliminate those traces. In this paper, we consider a timer abstraction that introduces a cyclic behaviour on abstract timers and we show how one can exclude cycles by imposing a strong fairness constraint on the abstract model. By employing the fact that the loop on the abstract timer is a self-loop, we render the strong fairness constraint into a weak fairness constraint and embed it into the verification algorithm. We implemented the algorithm in the DTSpin model checker and showed its efficiency on case studies. The same approach can be used for other data abstractions that introduce self-loops.