If you are an engineer who uses driving simulators to develop Advanced Driver Assistance Systems (ADAS), then the terminology “ground truth” means something specific to you. Ground truth representations define how your sensors and on-board systems interact with the virtual world, with calculated surfaces and objects. But if you are a technical manager, “ground truth” can have a double meaning.
At this junction in history, Advanced Driver Assistance Systems (ADAS) and other on-board electronic systems are not only a reality; they are a dominating presence within vehicle development programs. Sophisticated sensor-plus-intelligence features that have been on the concept board for decades have already been successfully introduced via high-end vehicles, and are now commonplace in city cars – e.g. Automatic Emergency Braking (AEB). ADAS-type sensors and information processing technologies are also on the forward edge of autonomous vehicle control – literally and figuratively.
Conducting comprehensive fail-safe tests with real drivers in prototype vehicles introduces additional costs and potential hazards (injury and/or damage to assets) for vehicle development programs. Since performance verification of vehicle electronic control systems and on-board active safety technologies is a must, vehicle manufacturers are obliged to seek efficient ways to make it happen. Vehicle dynamics class Driver-in-the-Loop (DIL) simulators are emerging to satisfy this need.
Modern motor vehicles are digital and electromechanical marvels, incorporating ever-greater numbers of increasingly complex, computer-controlled components and subsystems. As a result, ‘systems integration’ has become an integral part of vehicle development programs, as vehicle manufacturers bring aboard new technologies from their supplier networks.
Technical managers are wondering how to jump start their vehicle development programs while leveraging their existing Tier 1 (and other) resources. At one extreme, one might ask if there is an effective way to verify the integration of supplied ‘black box’ systems; at the other extreme one might ask how to efficiently explore thousands of tuning parameters (and whether the ability to do so is a blessing or a curse!).
I time traveled last week and I have returned to describe the wonders of the future. My time machine was not a DeLorean (as in the Back to the Future film series), it was an autonomous, electrically powered pod, a silver and blue device that hummed along calmly, and deposited me at earth coordinates 44°53'31.6"N 0°33'58.4"W, the doorstep of the Intelligent Transport Systems (ITS) World Congress. I’m not sure in what year I landed. I would guess 2020 or thereabouts. I did not time travel alone. There were many others.
Today’s automobiles have as much in common with advanced consumer electronics as they do transportation. In fact, according to Car & Driver magazine, your car is the most advanced electronic device you own, with high-end luxury vehicles typically sporting more than 100 electronic control units (ECUs).
Safe and reliable operation of the vehicle is, of course, the primary objective during electronic systems confirmation testing - But there is more at stake here for automotive manufacturers. A subset of a vehicle’s computing power directly affects its drivability – its vehicle dynamics fingerprint and subjective character, which are crucial to brand identity and value perception. Over many years, companies like Ford and BMW (“Sheer Driving Pleasure”) have anchored their automotive brand identity on the way their cars drive. For these and many other manufacturers it is crucial that the myriad electronic systems placed between the driver and the road do not detract from the actual driving experience.
Advanced Driver Assistance Systems (ADAS) implementations are on the rise as vehicle manufacturers continue to introduce greater levels of supplemental control into the driving experience, with greater safety and convenience as the goals.
Early testing in a digital environment is just as applicable to ADAS development as it is to everything else on board – perhaps even more so. Consider the case of introducing a new ADAS function such as a lane departure warning system. The system may be unfamiliar to drivers, thus making their reactions harder to predict. It is certainly desirable to connect real drivers with such a system as early as possible. How can this be achieved?
Traditionally, subjective evaluation of steering system performance has required the use of real test vehicles on real proving grounds. This stands to reason because the tactile feedback of a steering wheel is such an important part of the subjective driving experience, and historically speaking, the simulation of steering feeling has been one of the most challenging performance tasks for driving simulators.
At the heart of any engineering class Driver-in-the-Loop (DIL) simulator is a vehicle physics model, a mathematical representation of the systems of the driven car.
Different DIL experiments often require different types of vehicle models. Sometimes this is due to the desire for computational efficiency – i.e. Real-time calculation resources can be allocated to the vehicle subsystem that is of interest for the particular DIL experiment.
It can be agreed without debate that tuning and developing cars requires both subjective and objective assessments. Subjective information typically comes from experienced drivers with an ability to meaningfully describe vehicle behavior. Objective information typically comes from off-line simulations, on-car sensors and competent data acquisition and analysis. How to weigh all this information and mold it into a vehicle identity is another (big) topic altogether – so for the moment, let’s focus on the information itself, and more specifically address this question: Is it possible to gather both subjective and objective measurements in driving simulators? The answer is a qualified “yes.”
