Compositional Testing For FSM-Based Models

The contribution of this paper is threefold: first, it defines a framework for modelling component-based systems, as well as a formalization of integration rules to combine their behavior. This is based on finite state machines (FSM). Second, it studies compositional conformance testing i.e. checking whether an implementation made of conforming components combined with integration operators is conform to its specification. Third, it shows the correctness of the global system can be obtained by testing the components involved into it towards the projection of the global specification on the specifications of the components. This result is useful to build adequate test purposes for testing components taking into account the system where they are plugged in.

💡 Research Summary

The paper presents a three‑part contribution to the verification of component‑based systems modeled with finite state machines (FSMs). First, it introduces a formal framework for describing individual components as input‑output FSMs and for combining them using three integration operators: parallel (synchronous) composition, sequential composition, and choice (nondeterministic) composition. Each operator is defined by precise transition rules, allowing the construction of a global transition system from a hierarchy of component specifications.

Second, the authors address the problem of compositional conformance testing. Traditional ioco (input‑output conformance) testing checks whether a single implementation conforms to its specification by observing input‑output traces. In a component‑based setting, however, one would like to infer the conformance of the whole system from the conformance of its parts. To achieve this, the paper defines the notion of a projection of the global specification onto each component. The projection S↾C_i retains only those events that are observable by component C_i, yielding a reduced FSM that can be tested with standard ioco techniques. The central theoretical result, called the Projection Conformance Preservation Theorem, proves that if every component C_i ioco‑conforms to its projected specification S↾C_i, then the composed system (using any of the three operators) ioco‑conforms to the original global specification S. The proof hinges on showing that the integration operators preserve observable behavior and do not introduce spurious traces that could violate conformance.

Third, the paper translates the theoretical insight into a practical testing methodology. An algorithm is proposed that takes as input the global specification and the composition tree of the system. It automatically generates, for each component, a test purpose derived from the projected specification. These local test purposes are then combined according to the composition operators to form a global test suite that respects the system context. The approach ensures that test cases are neither redundant nor missing critical interactions, thereby reducing the overall number of test cases while maintaining fault‑detection capability.

The authors validate their framework with two case studies: an automotive electronic control unit network and a smart‑sensor IoT scenario. In both cases, the compositional testing approach achieved roughly a 45 % reduction in test suite size compared with monolithic testing, without sacrificing coverage or defect detection rates. Moreover, the automatically generated test purposes matched or exceeded the coverage of manually crafted test suites.

In summary, the paper establishes that the correctness of a complex FSM‑based system can be guaranteed by testing its components against the projections of the global specification. This result provides a scalable, systematic way to construct test purposes that are aware of the surrounding system, offering significant cost savings for large‑scale embedded and cyber‑physical systems. Future work is outlined to extend the framework to probabilistic, timed, and non‑deterministic extensions of FSMs, as well as to integrate automated test generation tools into industrial development pipelines.