A Method for Reducing Simulation Timing Deviation in QEMU-Based Virtual ECU

As support for a wider range of functions and services of vehicles, such as driver assistance systems, autonomous driving, and OTA services, has increased in recent years, the complexity of software embedded in vehicles has become increasingly complex. These changes are driving demand for simulation technologies that can help reduce development and validation time while ensuring vehicle safety and reliability. Among them, electronic control unit virtualization technology is steadily being researched and commercialized by many companies because it can consider the characteristics of ECUs from the development and validation stages of in-vehicle embedded software. In this paper, we discuss the virtualization method of STM32F407ZGT using QEMU as part of ECU virtualization technology. Furthermore, we propose a technique to reduce the simulation error by measuring the running time of the physical ECU and adjusting the unit time in the AUTOSAR OS on the virtual ECU. This is to reduce the simulation error caused by the dependence on the host computer environment and the generation of overhead in the case of highly computational in-vehicle embedded software, which have been pointed out as problems in ECU virtualization using QEMU. The experimental results of the proposed technique show that the simulation timing error rate of the virtual ECU compared to the physical ECU is reduced from an average of 2.40% to 0.02%. This shows that, just by reflecting the operational characteristics of the physical ECU in the QEMU-based virtual ECU, it is possible to improve the precision by reducing the timing error of the simulation.