Low-latency graphics bring autonomous vehicles a step closer

The upsurge of interest in autonomous and ADAS systems has brought a corresponding increase in the demand for testing in order to confirm their behaviour and validate their response to the unpredictable conditions encountered on public roads. The factors limiting virtual testing in conventional simulation environments have been relatively low graphics quality and poor latency, resulting in a system too slow to use in an emergency manoeuvre.

While most OEMs and Tier 1 suppliers have successfully adopted a model based engineering process for the development of their ADAS control systems, conventional graphical simulation environments cannot respond fast enough to cope with the highly dynamic manoeuvres experienced when testing safety systems. rFpro’s solution was developed from the ground up to deliver driving simulation for vehicle dynamics applications and allows OEMs to re-introduce professional human test drivers into the model based development process.

Apart from improved speed of response, rFpro’s software provides much higher quality graphics: essential when testing an autonomous system’s ability to distinguish between features with a similar appearance. Real-world lighting conditions create complex shadows that can confuse ADAS camera systems.

Another key aspect of ADAS development is interaction with the driver of the vehicle, who may panic or behave unpredictably during emergency avoidance, instinctively grabbing the steering wheel or fighting the system. Evaluating this type of interaction with a simulator also requires high graphics quality to provide convincing realism and low latency for lag-free feedback to the driver.

“Autonomous systems must make split-second decisions, just like the drivers they protect, so they can only be tested effectively with graphics software that refreshes quickly enough,” said Chris Hoyle, Technical Director, rFpro. “Our system delivers the graphics up to 100ms faster than typical 3D engineering graphics, which is equivalent to a car length travelled at highway speeds. This can make the difference between reacting before or after an impact. Winter sunlight, low in the sky, can generate shadows from roadside objects such as crash barrier uprights, which the system then confuses with lane markings. This type of situation can be investigated and corrected by testing with our simulation software. Solutions from PreScan and IPG allow our graphics to be fed straight to the control systems being tested, improving the quality of information delivered compared to the traditional approach of pointing a camera at a computer generated display.”

“Testing autonomous systems without a DIL will inevitably lead to inaccurate results,” continued Hoyle. “A computer will respond to an emergency situation consistently linearly, a human does the opposite. So it is essential to test these systems with a driver’s response in a safe environment before testing on public roads. In an ironic twist, the same software developed to make DIL simulators effective in allowing a human driver to interact with a virtual car, is now helping to develop ADAS technologies that take the driver ‘out’ of the loop in an emergency. The benefits are very similar: the HMI can be explored repeatedly under controlled conditions, including any changes required, in complete safety.”