A Genetic Algorithm for Goal-Conflict Identification

Abstract

Goal-conflict analysis has been widely used as an abstraction for risk analysis in goal-oriented requirements engineering approaches. In this context, where the expected behaviour of the system-to-be is captured in terms of domain properties and goals, identifying combinations of circumstances that may make the goals diverge, i.e., not to be satisfied as a whole, is of most importance. Various approaches have been proposed in order to automatically identify boundary conditions, i.e., formulas capturing goal-divergent situations, but they either apply only to some specific goal expressions, or are affected by scalability issues that make them applicable only to relatively small specifications. In this paper, we present a novel approach to automatically identify boundary conditions, using evolutionary computation. More precisely, we develop a genetic algorithm that, given the LTL formulation of the domain properties and the goals, it searches for formulas that capture divergences in the specification. We exploit a modern LTL satisfiability checker to successfully guide our genetic algorithm to the solutions. We assess our technique on a set of case studies, and show that our genetic algorithm is able to find boundary conditions that cannot be generated by related approaches, and is able to efficiently scale to LTL specifications that other approaches are unable to deal with.