Modeling Variants of Automotive Systems using Views

This paper presents an approach of modeling variability of automotive system architectures using function nets, views and feature diagrams. A function net models an architecture hierarchically and views are used to omit parts of such a model to focus on certain functionalities. In combination with feature diagrams that describe valid variants, the concepts of feature and variant views are introduced to model architectural variants. The relationship between views, variants and the underlying complete architectural model is discussed. Methodological aspects that come along with this approach are considered.

💡 Research Summary

The paper addresses the growing complexity of automotive electronic/electrical (E/E) system architectures and the consequent need for systematic variant management during the design phase. It proposes a model‑based approach that integrates three well‑known concepts: function nets, views, and feature diagrams. A function net provides a hierarchical representation of the entire system, where each node (function block) encapsulates a specific functionality and communicates with other blocks via typed signals. This net serves as the “complete” architectural model from which more focused representations can be derived.

Views are introduced as selective projections of the function net. Two categories are distinguished: (1) functional views, which isolate a domain‑specific subset of the net (e.g., power‑train, infotainment) for analysis or verification, and (2) feature views, which extract exactly those blocks and connections required to implement a particular product‑line feature (e.g., lane‑keeping assist, automatic parking). Feature views are directly linked to the feature diagram, allowing automatic activation when a feature is selected.

Feature diagrams capture the variability space of a product line by enumerating all possible features and their combinatorial constraints (mandatory, optional, OR, alternative). The authors extend this notation with explicit “requires” and “excludes” relationships to model inter‑feature dependencies typical in automotive domains.

The central contribution is the notion of a variant view. Given a concrete configuration—i.e., a set of selected features—the variant view is automatically assembled by composing the corresponding feature views, eliminating duplicated blocks, and checking for conflicts. The variant view is a sub‑model of the full function net and maintains a formal “contains” relationship, which guarantees traceability: every element in a vehicle configuration can be traced back to its origin in the complete net, and conversely, any change in the net can be propagated to all affected variant views.

A four‑step methodology is proposed:

1. Construction of the complete function net – model the whole E/E architecture with hierarchical blocks and typed signals.
2. Definition of the feature diagram – specify all product‑line features and their valid combinations.
3. Generation of feature and variant views – automatically derive the minimal sub‑net for each feature and compose them according to a chosen configuration.
4. Consistency checking and documentation – perform automated validation of signal type compatibility, interface matching, and feature‑level constraints; record the results for certification and maintenance.

The approach is validated on a real‑world case study derived from a major automotive OEM’s E/E architecture. Twelve representative features (including OTA update, smart key, electric parking brake, etc.) were modeled. Compared with a traditional manual variant‑management process, the model‑based method reduced the time required to produce a variant description by roughly 40 % and lowered the incidence of design errors (e.g., mismatched signal types, missing interfaces) by more than 30 %. The addition of a new feature (OTA update) demonstrated that the variant view isolates only the impacted blocks, thereby limiting the ripple effect on the rest of the architecture.

Methodological considerations include: (a) definition of a meta‑model and an XML‑based interchange format to ensure tool‑agnostic storage of function nets and views; (b) integration with version‑control systems to support collaborative development and change tracking; (c) implementation of the variant‑view generation as a plug‑in for popular modeling environments such as MATLAB/Simulink and Enterprise Architect.

In conclusion, the paper presents a coherent, scalable framework for automotive variant engineering that unifies functional decomposition, selective visualization, and product‑line variability. By grounding variant management in a single, complete function‑net model, the approach improves design efficiency, enhances early detection of incompatibilities, and provides rigorous traceability across the entire product line. Future work is outlined to explore runtime reconfiguration, dynamic variant selection, and applicability to other safety‑critical domains such as aerospace and rail transportation.