[Paper Review] Towards Controlling Refinements of Statecharts

Conner Hansen, Eugene Syriani+1 more|arXiv (Cornell University)|Mar 25, 2015

Model-Driven Software Engineering Techniques5 references3 citations

TL;DR

This paper proposes a formal refinement relation for Statecharts that ensures external behavior preservation during incremental development by enforcing strict bounds on allowable modifications. By restricting refinements to additive changes—such as adding transitions and states within 'virtual' states—while preserving events, guards, and structural connectivity, the approach guarantees backward compatibility and predictability, enabling safe reuse and evolution of models without breaking existing assumptions or behavior.

ABSTRACT

In incremental development strategies, modelers frequently refine Statecharts models to satisfy requirements and changes. Although several solutions exist to the problem of Statecharts refinement, they provide such levels of freedom that a statechart cannot make assumptions or guarantees about its future structure. In this paper, we propose a set of bounding rules to limit the allowable Statecharts refinement operations such that certain assumptions will hold.

Motivation & Objective

• To address the lack of control in existing Statecharts refinement approaches, which allow arbitrary modifications that break behavioral assumptions.
• To enable safe, incremental model evolution by enforcing constraints that preserve the external behavior of original statecharts.
• To provide a formal mechanism for refinement that supports reusability and maintenance while aligning with software engineering principles like the open-closed and Liskov substitution principles.
• To define a bounded, realistic set of rules that limit refinement operations to only those that maintain the original model's behavioral integrity.
• To introduce 'virtual' states as a structural construct to localize and control where new transitions and states may be added during refinement.

Proposed method

• Define a refinement relation R between an original statechart SC₀ and a refined statechart SCᵣ using inverse surjection to ensure each element in SCᵣ maps to exactly one element in SC₀.
• Enforce event and variable inclusion (E₀ ⊆ Eᵣ and V₀ ⊆ Vᵣ) to preserve the external interface of the original model.
• Enforce guard inclusion by requiring that every guard in a refined transition must correspond to a guard in the original transition, preserving preconditions.
• Enforce structural inclusion via LTS (Labeled Transition System) semantics: any transition path in SC₀ must be realizable in SCᵣ, either directly or through intermediate steps, preserving connectivity.
• Restrict new transitions and states to 'virtual' states—specialized container states that allow controlled extension without altering the original statechart’s core structure.
• Prohibit removal of states, transitions, or events; disallow modification of source/target states or event triggers; allow only conjunctions (not disjunctions) in guards to avoid weakening preconditions.

Experimental results

Research questions

• RQ1 How can Statecharts refinement be controlled to preserve the external behavior of the original model during incremental development?
• RQ2 What formal constraints are necessary to ensure that refined models remain predictable and maintain backward compatibility with original assumptions?
• RQ3 In what ways can refinement be restricted to allow safe extension without violating behavioral guarantees or introducing unintended side effects?
• RQ4 How can the concept of 'virtual' states be leveraged to localize and control model evolution while preserving structural integrity?
• RQ5 What are the minimal yet effective rules that balance developmental freedom with behavioral safety in Statecharts refinement?

Key findings

• The proposed refinement relation successfully preserves the external behavior of the original Statechart by ensuring all original events, variables, and transition paths are maintained in the refined model.
• The use of inverse surjection and structural inclusion via LTS semantics guarantees that any behavior observable in the original model remains realizable in the refined version.
• Restricting new transitions and states to 'virtual' states enables controlled extension while preventing unintended structural changes that could break assumptions.
• The approach prevents harmful modifications such as event or guard changes, transition removals, or source/target reassignments, which are common in existing tools like UML and Rhapsody.
• Empirical evaluation on an incremental development example confirmed that the original behavior is preserved while still allowing extensive, safe extensions.
• The method outperforms existing approaches (e.g., UML, Rhapsody, μ-Charts) in ensuring behavioral predictability by enforcing stricter, behavior-preserving constraints.