A search-based technique for testing from extended finite state machine model

Abstract

Extended finite state machines (EFSMs), and languages such as state-charts that are similar to EFSMs, are widely used to model state-based systems. When testing from an EFSM M it is common to aim to produce a set of test sequences (input sequences) that satisfies a test criterion that relates to the transition paths (TPs) of M that are executed by the test sequences. For example, we might require that the set of TPs triggered includes all of the transitions of M. One approach to generating such a set of test sequences is to split the problem into two stages: choosing a set of TPs that achieves the test criterion and then producing test sequences to trigger these TPs. However, the EFSM may contain infeasible TPs and the problem of generating a test sequence to trigger a given feasible TP (FTP) is generally uncomputable. In this paper we present a search-based approach that uses two techniques: (1) A TP fitness metric based on our previous work that estimates the feasibility of a given transition path; and (2) A fitness function to guide the search for a test sequence to trigger a given FTP. We evaluated our approach on five EFSMs: A simple in-flight safety system; a class II transport protocol; a lift system; an ATM; and the Inres initiator. In the experiments the proposed approach successfully tested approximately 96.75 % of the transitions and the proposed test sequence generation technique triggered all of the generated FTPs.