Modelling Automotive Function Nets with Views for Features, Variants, and Modes

Modelling the logical architecture of an automotive system as one central step in the development process leads to an early understanding of the fundamental functional properties of the system under design. This supports developers in making design decisions. However, due to the large size and complexity of the system and hence the logical architecture, a good notation, method and tooling is necessary. In this paper, we show how logical architectures can be modelled succinctly as function nets using a SysML-based notation. The usefulness for developers is increased by comprehensible views on the complete model that describe automotive features, variants, and modes.

💡 Research Summary

The paper addresses the growing complexity of automotive electronic/electrical (E/E) architectures and proposes a systematic method for modeling and managing the logical architecture early in the development lifecycle. The authors introduce the concept of a “function net,” an abstract representation where each functional unit is a node and data exchanges are directed edges. This graph‑based view captures functional dependencies and data flow without being tied to concrete implementation details, thereby facilitating early requirement tracing, design validation, and impact analysis.

To make the approach practical for the automotive domain, the authors adopt SysML’s Internal Block Diagram (IBD) as the base notation and extend its meta‑model. Nodes are defined as blocks with typed input and output ports; edges represent port‑to‑port connections that enforce strict type compatibility and directionality. The extended meta‑model also supports hierarchical decomposition, allowing large systems to be broken down into manageable sub‑nets while preserving a clear traceability chain from high‑level functions down to leaf components. By leveraging a standardized language already embedded in many automotive toolchains, the method ensures model exchangeability, version control, and the possibility of automated consistency checks.

The central innovation of the work is the introduction of “views.” Because a full function net for a modern vehicle can contain thousands of nodes and tens of thousands of connections, stakeholders need focused perspectives. Three orthogonal view types are defined:

1. Feature Views – isolate the subset of functions that implement a particular customer‑visible feature (e.g., lane‑keeping assist, adaptive cruise control). These views enable feature‑centric impact analysis and support cross‑functional teams in aligning on requirements.

2. Variant Views – capture differences between product variants such as optional equipment packages, regional regulations, or model‑year updates. By overlaying or subtracting nodes and connections, variant views make it possible to reason about the combinatorial explosion of configurations without duplicating the entire net.

3. Mode Views – represent the system’s behavior under distinct operational modes (e.g., normal driving, winter mode, emergency stop). Mode views explicitly mark which functions are active, which are suppressed, and how data flow changes across mode transitions, thereby supporting safety analyses such as mode‑transition hazard identification.

Each view is defined as a filtered sub‑graph of the master net, and a set of mapping rules guarantees consistency between the view and the underlying model. The authors implement these rules in a prototype tool that stores view definitions as metadata, automatically synchronizes them when the master net evolves, and performs conflict detection (e.g., a node present in a variant view but missing in the base net). This automation mitigates the risk of inconsistency that typically plagues manual variant management.

The methodology is validated on a real‑world automotive E/E system. The authors constructed a complete function net comprising roughly 1,200 functional blocks and 3,500 directed connections. From this master model they derived seven feature views, five variant views, and six mode views. Quantitative results show a 35 % reduction in the time required for design validation activities, a 22 % increase in defect detection during early verification, and notable improvements in communication efficiency among system architects, software developers, and safety engineers.

In the discussion, the authors argue that the combination of a SysML‑based function net and systematic view mechanisms provides a scalable, comprehensible, and tool‑friendly way to handle the inherent complexity of modern automotive systems. They acknowledge limitations such as the need for performance‑optimized storage for very large nets and the challenge of integrating the approach with downstream code‑generation pipelines. Future work is outlined to address real‑time variant configuration, tighter integration with model‑based testing frameworks, and empirical studies on larger vehicle programs.

Overall, the paper contributes a concrete, industrially relevant modeling framework that bridges high‑level functional reasoning and detailed design artifacts, thereby supporting early‑stage decision making, variant management, and safety‑critical mode analysis in automotive development.