Sensor fusion—the combining of measurement data from many sensors to drive outcomes—will be necessary, and will require synchronisation, high-power processing, and the continued evolution of the sensors themselves.
Next year the Audi A8 will be the world’s first production car to offer Level 3 autonomy - equipped with six cameras, five radar devices, one lidar device, and 12 ultrasonic sensors.
Ultimately, the goal of processing sensor data is to create a fail-safe representation of the environment surrounding the car in a way that can be fed into decision-making algorithms and that can keep costs down.
ADAS processing capabilities are based on multiple isolated control units; however, sensor fusion is driving demand, among some manufacturers, for a singular centralised processor. But that can be costly and require vast amount of processing, as a result, other manufacturers are looking at a more distributed architectural design.
Among the most significant challenges for manufacturers is choosing the right software, and testing it against an infinite number of real-world scenarios.
“All of which means a software-defined tester will be critical in keeping up with that evolution,” explains Jeff Phillips, Head of Automotive Marketing at National Instruments (NI). “The semiconductor content will only continue to increase and as the technology improves so the ability to quickly reconfigure testers will be critical. Flexibility through software will be the key.
“Increasing the efficiency in software development will be integral to the autonomous driving revolution,” says Phillips.
“To be frank, a lot is changing in the testing environment. One of the biggest challenges with ADAS is that the actual test environment has got to be able to handle a much wider array of challenges. In the past we focused on a vehicle’s physical characteristics, today we need to consider electronic measurement and the role of the software in running ECUs.”
Software testing involves a very different testing methodology, according to Phillips. “When we talk about autonomous vehicles much of the software that will run on the ECU will have been written by neural networks or other non-human systems – so the tracking of changes back to source code is more difficult, as is validating the software as it’s always updating.
“Today, we talk more about predictive rather than deterministic testing. In the past test was based on a simple ‘yes’ or ‘no’ response. ADAS is not deterministic and there’s not always going to be a clear answer, much of what we will have to do will be based on interpretation.
“As a result we’ll need more software engineers, but also full blown software architects.”
While some manufacturers favour consolidating ECU functionalities with the focus on sensor fusion, infotainment and those functions requiring less data – such as steering, air conditioning etc., others favour a more distributed architecture.
“Consolidation of the ECU can be costly and will require immense processing capabilities. There’s also a greater risk that with fewer ECUs the vehicle could be more vulnerable to hacking and security breaches.
“There are also concerns that the distributed nature of production means that there is global accessibility to data, which could leave systems vulnerable.”
One of the big issues confronting the industry is a lack of standards.
“Car companies are going in different directions,” says Phillips. “With different sensors being used, decisions are likely to vary from device to device. But I believe that when it comes to probabilistic testing we are reaching an inflection point. There will need to be mandates and agreements at government level.”
When it comes to test car manufacturers are engaging with companies like National Instruments very differently.
“They no longer simply go out and buy a tester, they engage with us to design their own tester platforms that they can own and develop. That’s a massive change. Because of the costs associated with test they are looking to build up their own tester infrastructure. They are making their own decisions and looking to the likes of NI to provide solutions.
“In truth everybody knows the general direction, but most don’t now the address. Too many are still scrambling around and there remain a lot of unknowns for engineers to address.
