Experimental Evaluation of Timed Finite State Machine Based Test Derivation

Finite State Machines (FSMs) are widely used for deriving tests with guaranteed fault coverage for various kinds of discrete and hybrid systems such as telecommunication protocols and services. A Timed FSM (TFSM) seems to be a natural extension of an FSM annotated with timed variables that allows to adequately describe a system under test for which time aspects are important. However, most of existing test derivation methods for timed automata with several clocks return very long test suites. We propose to consider FSMs with a single clock for deriving test suites with guaranteed fault coverage and this can be a compromise between the model complexity and adequacy. A test is derived based on FSM abstraction of a given TFSM and is complete with respect to an established fault model. The experiments were performed for randomly generated TFSMs as well as for TFSMs which are used for service description and they clearly show that a proposed method returns high quality tests with the length that is close to those returned by classical FSM based methods. For service TFSMs we also evaluate the test suite length for different fault domains, the length of preset distinguishing and homing sequences which are widely used for minimizing FSM based test suites with guaranteed fault coverage.