Parallelisation of Virtual ECUs for Vehicle Simulation - Design and Implementation of FMI 3.0 and FMI-LS-BUS Based CAN Communication for Scalable Automotive System Simulation

Abstract

Calibration of complex propulsion systems in commercial vehicles increasingly relies on large-scale Co-Simulation of Virtual Electronic Control Units (vECUs). However, existing simulation pipelines are constrained by non-parallel execution and dependencies on host-configured virtual CAN devices, limiting scalability and portability. This thesis presents a scalable, parallelisable Co-Simulation architecture for vECUs using the Functional Mock-up Interface (FMI) 3.0 standard and its FMI-LS-BUS extension for network communication. We developed a proof-of-concept system that exports vECU models as FMI 3.0 Co-Simulation FMUs with Low-Cut FMI-LS-BUS support, enabling bus-oriented communication without manual signal-level mapping. The implementation bridges Volvo TTI’s internal vECU CAN driver with the FMI interface, utilising dynamic generation to derive CAN network terminals from vehicle configuration artefacts. The system leverages the SIL Kit middleware by Vector for inter-process communication, replacing OS-dependent virtual devices with platform-independent messaging. Critical contributions include modifications to the SIL Kit FMU Importer to correctly handle serialised sequences of multiple CAN operations within single FMI binary variables. Evaluation results demonstrate that the prototype successfully preserves the logical semantics of CAN communication across multiple parallel processes. Benchmarks with up to 100 concurrent FMU participants show linear scalability in per-process runtime when sufficient processor cores are available, with approximately 25% overhead relative to non-communicating simulations in the same benchmark configuration. The automated network mapping eliminates the need for thousands of manual signal connections as participant count grows. While the prototype operates at the logical frame-exchange level without modelling physical bus timing, the results confirm that FMI-LS-BUS is a viable foundation for scalable automotive system simulation.