Automated Assessment of E/E-Architecture Variants Using an Integrated Model- and Simulation-Based Approach
To be published on April 2, 2019 by SAE International in United States
Due to the continuously increasing complexity of automotive electric/electronic architectures (EEAs), model-based systems engineering principles became state-of-the-art for designing such heterogeneous systems. However, current Architecture Description Languages (ADLs), their system-design and analysis tools as well as frameworks for simulation-based analysis of EEA models often are not fully integrated in a single design process but usually require error-prone import/export processes, especially when considering distributed collaboration, from EEA data models to external analysis frameworks and vice versa. Particularly, this barriers the efficient assessment and comparison of distinct architecture variants regarding certain non-functional properties. Moreover, simulation-based analysis of the latter intended for EEA assessments in early concept phases necessitates backtracking capabilities to allow iterative model adaptations. In this paper, we present a novel approach for evaluating EEAs dynamically in different use-cases and comparing their architecture variants. We contrived a generic concept to integrate actor-oriented simulation and analysis tools with EEA system-design tools based on a previously developed approach for synthesizing and executing a cross-domain simulation model out of static EEA descriptions in PREEvision. Additionally, our concept comprises the design model’s execution in a loop to allow iterative modifications of customer specific metrics as well as the automated assessment of their impact on the model. Furthermore, we added a variant-sensitive synthesis mode to analyze and compare distinct architecture variants and find the best suited one according to a considered use-case. Therefore, intermediate simulation results and the UUIDs of their source model artifacts are fed back and stored coherently in a simulation result registry. This facilitates to link the synthesized simulation model artifacts with the original EEA model artifacts and thus enables backtracking and the usage of the gained data in further analysis metrics. To proof this concept, we focus on the evaluation and comparison of two architecture variants (a centralized and a distributed hardware topology) by means of the incurred communication latencies of several activity chains.