Integration of Model Based System Engineering (MBSE) environment with Multi-Disciplinary Optimization (MDO) approach applied to the Design of Automotive Electronic Hardware Architecture

In modern vehicles, the ongoing increase in capability, connectivity, and automation has led to a substantial rise in both the number and complexity of Electronic Control Units (ECUs). Balancing this growing complexity within strict cost, weight, and safety boundaries has become a major challenge for automotive manufacturers. Traditional document-based engineering approaches are no longer able to ensure consistency between requirements, design, and implementation, which motivates the use of more integrated and model-driven approaches capable of managing multidisciplinary trade-offs from the earliest phases of system development. The objective of this thesis is to develop and validate a framework for the preliminary design and optimization of automotive electronic architectures. In particular, the work addresses the question of how to allocate vehicle functions to ECUs in order to minimize cost and weight while ensuring compliance with performance and communication constraints. Furthermore, the framework aims to facilitate early phase decision making by supporting engineers in the exploration and identification of feasible designs across multiple architectural alternatives, even under complex design constraints. To achieve this goal, the thesis combines Model-Based Systems Engineering (MBSE) and Multidisciplinary Design Optimization (MDO). MBSE was implemented using Cameo Systems Modeler with the MagicGrid methodology to capture requirements, system functions, and architectures in SysML. The resulting model was then converted into structured CSV tables representing functions, computational loads, and communication signals. The optimization environment was developed in Python using the GEMSEO library, supported by a dedicated ECU database built on Intel and AMD FPGA/SoC devices. Two types of optimization problems were designed. In the single-objective case, both continuous and discrete formulations were applied: the continuous cost model was derived through curve fitting in MATLAB, incorporating computational performance and bus connectivity, while the discrete model relied on catalog-based ECU selection. In the multi-objective case, cost and weight were minimized simultaneously, producing Pareto fronts that describe trade-offs in the allocation process. In all formulations, constraints on computational load, bus capacity and maximum number of function for each ECU were enforced to guarantee the validity of the optimized architectures. The results of this thesis demonstrate that it is indeed possible to integrate MBSE and MDO environments, thereby improving both traceability across system representations and the quality of early decision-making in the design process. The optimized architectures obtained through the proposed framework show promising results in terms of cost reduction and mass efficiency, confirming the effectiveness of the approach. Nevertheless, future work should aim to extend the ECU database to provide greater variety and broader options for architectural allocation, as well as to introduce additional optimization disciplines such as power consumption, safety, and ECU placement. More broadly, the framework demonstrates the potential of combining system modeling and optimization to address the challenges of complex automotive electronics, supporting robust and informed decision-making from the earliest design stages.