Software-Defined Architectures Disrupting Auto Industry

Over the course of more than 100 years, the car industry has refined and optimized what we now recognize as a vehicle. But it generally still has four wheels and an engine, and even the move to electric motors hasn’t introduced that much change in appearance and function.

Consumers now ask new questions, of course, such as “How safe is an autonomous car?”, “Is it really zero-emissions?” and “Has it been designed and built in an environmentally ethical way?” They also want to know that the data they share with and through their vehicle is as safe as their family riding inside.

These questions are based on the disruptive nature of technology. It is no coincidence that the new entrants into the automotive industry do not look like traditional automotive manufacturers. The new platforms they produce will not be like traditional vehicles.

Modern consumer products all look the same until you turn them on, and then you recognize your own phone (and it recognizes you). In a similar way, the future vehicle will provide the scope for customization and personalization on a level never seen before. This scope will be enabled by software and the underlying hardware will be software-defined.

Recent supply issues have highlighted the importance of semiconductors to the automotive industry. The differentiating features and services provided by connected vehicles will continue to be enabled by semiconductor technology. Software is a fundamental part of the overall experience, but it must be complemented with the best hardware.

Form Follows Function, Silicon Follows Software

There are many challenges that this hardware/software relationship must still face and overcome. Replacing a vehicle’s driver will take millions of lines of code, and the car must continue to offer its occupants a familiar experience.

To address this and to optimize for space, weight and cost, functions will become more centralized. Virtualization will be used to run multiple instances in cross-domain platforms. Functions that once required hundreds of individual ECUs will become fully integrated on the same hardware running in a robust, safe and secure software environment.

This level of complexity can only work if there is an optimal balance between hardware and software.

In many cases, the conventional and sequential “if, then, else” programming methodology will be replaced with Artificial Intelligence and Machine Learning technologies. Once trained, using huge amounts of data that has been categorized for quality, relevance and accuracy, AI/ML will operate faster and more safely than the driver’s brain.

How does this level of change impact semiconductor vendors? Automakers will need to re-evaluate the embedded environment. They must discover how highly optimized hardware accelerators will be used to handle the extreme processing loads that the complex mathematical operations and matrix multiplications will demand of the hardware.

There will be an unprecedented amount of real-time data to process, coming from the huge number of sensors integrated into the car.

Naturally, the software running on these optimized hardware platforms will require constant maintenance, demanding secure over-the-air (OTA) updates. While some of the more disruptive automakers have already started implementing this, others have yet to fully grasp its significance. Semiconductor technologies support this through actively managing memory partitions and processes that guarantee stability even during a live update.

Daniel Weyl NXP Semiconductors

This is only possible through the development of high-performance, highly reliable semiconductor technology.

Daniel Weyl (pictured, left) is vice president-R&D software at NXP Semiconductors and leads the global Software & Tools R&D organization of NXP Automotive Processing.