The goal of the paper is to describe how the non-dominated sorting genetic algorithm III (NSGA-III) algorithm generates test suites from extended finite-state machines (EFSMs). Each generated test case represents a feasible transition path on the EFSM. The ability to deal with feasibility is one advantage of using EFSMs instead of FSMs, thanks to the use of context variables and guards associated with transitions.

NSGA-III is a genetic algorithm that has been applied to multi-objective problems. In its application for test case generation, three objectives are defined: “one objective for estimating the feasibility of each path, one objective for covering all transitions and minimizing the necessary paths, and the third objective [is the] minimization of similarities between tests.”

Chromosomes encode variable-length sets of paths where each gene is a list of integer numbers representing transitions. The solutions of the genetic algorithm are “those chromosomes that have only feasible paths and cover all transitions.”

An experiment was run with two different EFSMs, comparing the performance of the proposed algorithm with a previous proposal based on NSGA-II. The computation time was lower and the success rate (feasible and full coverage solutions) was higher.

The paper is inspired by Kalaji et al. [1] and Asoudeh and Labiche [2], and is an evolution of two previous works by Ţurlea published in 2017 and 2018.

This clear and concise paper mainly addresses researchers in automated test case generation.