57 HIL testing | Deep insight E-Mobility Engineering | March/April 2025 “At the model-in-the-loop (MIL) up to SIL you want subjective feedback,” he says. “You can run offline simulation first to find the optimum for performance, energy consumption; then select the experience for those selected ranges. “For example, you can have the best lane-keeping algorithm, but when you drive it the control system is a poor experience, too aggressive; then you move from the objective data to the subjective feel.” “With an advanced model, you can test energy consumption, the kWh/km, but no matter how long you run offline simulations, sooner or later you’ll need to drive that powertrain. “If you don’t drive it you will need to do a lot of tuning later on, and that’s very time-consuming and costly, so what you want to do is drive the vehicle virtually, and assess the quality and realism of the drive of the subsystems as if they are physically mounted on the vehicle, so you need a proper vehicle model, a proper motion platform, that is as close as possible to reality. “The components are interacting with other components, so you need to experience the whole assembly of the vehicle.” “If you test a physical component, you can find it’s 18 months before you have the full vehicle and find out if that component being developed is good enough – that’s the real advantage to having the HIL in a real-time simulator. “There is no better test bench than a human being. You don’t know how the driver will behave. Different drivers will behave in totally different ways, for example, to avoid an obstacle – professional drivers, the less experienced, new drivers – but you still need to have the vehicle respond in a safe way and interact with sensors that help the driver,” he says. “This enhances the quality of the test; for example, on the Nuremberg ring there will be many runs, and every single one will be different and the powertrain response will be different as it will stress in a totally different way. What is the right one? All of them are right. “Our work starts much earlier in the design process, long before the physical hardware is available, so our initial approach is not in HIL but in the early X-in-the-loop, where only the models are available, so we form a basic model to a more complex, validated, physical testing, but still at the model level.” “Then, we move to software-in-theloop (SIL), where the exact specific software of the ECU is running at real time at 1 kHz for vehicle dynamics, updating every 1 ms. We can control all the sublevels of the simulation, so we make sure every calculation is performed in that 1 ms,” he says. “Control systems and test rigs run faster at 4-8 kHz, but those exchange information with the driving simulator or the real-time test system every 1 ms. This is a common understanding for automotive applications.” The SIL covers the ECU with the exact communication protocols, such as CAN or FlexRay. “At every level you can have a mix of models, and as the products mature you can integrate the HIL, ECUs and subsystems with the proper communication protocols. ECUs can be open and reprogrammed, or they can be black box and fixed, and this is a very strict HIL definition,” Baldari says. “Mechanical HIL is a further stage, where you can use test rigs to include physical components, brakes, steering, connected to the real-time simulation, so you can have different simulation models or drive those test systems, and this is mechanical HIL. “The very final stage is a full vehicle where some of the components are missing – the vehicle-in-the-loop – where you can run on the vehicle in real time or faster than real time to replace ECUs that are not available.” This can be used to predict physical issues by using data from physical sensors such as radar or from the suspension system to anticipate braking, or for a rear-wheel steering system to see how the system performs over time. The AutoHawk hardware-in-the-loop system (Image courtesy of VI-grade)
RkJQdWJsaXNoZXIy MjI2Mzk4