OPINION: Building platform software for ADAS

OPINION: Building platform software for ADAS

Sreenivasa Reddy Berahalli, senior software engineering manager, Magna Electronics, discusses the necessity to develop software platforms that can abstract ever-evolving hardware platforms

Advanced driver assistance system (ADAS) technologies are evolving at an unprecedented rate, with new and innovative features added to vehicles every model year. Competition is fierce as silicon manufacturers produce newer and more capable chips at an even faster pace. This creates an urgent need to design the ideal hardware and software platform to host these new features.

The emphasis has shifted to creating software platforms for these continually evolving hardware platforms. The combination of a multidimensional and competitive silicon industry, ever-changing hardware and software technologies, and inconsistent industry standards requires manufacturers to stay ahead of the curve.

Advancements in ADAS
ADAS features combine active and passive features to minimize human error. With human error cited as the cause of about 90% of vehicle accidents and fatalities, the benefits of ADAS are immeasurable. The technology is growing at an unprecedented rate, valued at US$24bn in 2018 and predicted to reach more than US$90bn by 2025. This growth is fueled, in part, by increased government regulations that require rearview cameras, automated braking systems and other technologies.

The challenge for manufacturers is finding the best hardware and software platforms to host these features at a cost-competitive price. Meanwhile, silicon manufacturers are fabricating new, more capable chips, creating robust industry competition. It’s great for business but adds to the burden of porting and validating this software onto new hardware. As such, the software industry emphasizes developing platforms that accommodate the evolving hardware.

Software development and cost concerns
Twenty years ago, most automotive costs were in mechanical engineering – the production of seats, brakes, transmissions and other essential vehicle features. Today, car manufacturers spend more resources on electronics and software; the best example is the growth of integrated circuits, or system-on-chips (SoC), including chip development for electronic vehicles. One major drawback is the lack of standardization due to differences in system architectures.

All manufacturers use advanced RISC machine (ARM) chips, which are processor cores that improve software portability across various platforms. At the same time, the unique hardware accelerators constitute the silicon manufacturer’s intellectual property, rendering the software unique for each SoC platform rather than hardware-independent.

Understandably, the competition is fierce among silicon industry leaders competing to achieve market share in the automotive high-performance ARM architecture-based silicon space. Key players in this space include Nvidia, Qualcomm, Texas Instruments, Intel (MobilEye), AMD (Xilinx), Renesas and Ambarella.

As the software options and capabilities grow, so does the price impact on manufacturers and consumers. “By 2030, the global automotive software and electronics market is expected to reach US$462bn, representing a 5.5% [compound annual growth rate] CAGR from 2019 to 2030,” according to McKinsey & Company projections.

The combination of robust industry competition and exploding prices creates an environment where compatibility and continued growth are challenges. In the early days of the computer industry, Intel dominated the hardware scene, with Microsoft the clear leader in software. Software development evolved more easily for a single or standard set of hardware CPUs on Microsoft or Linux OS. In today’s cost-competitive ADAS space, the focus appears to be on cost-optimized, custom-designed and specific-purpose-developed ECUs and unique software for each ECU embedded system.

Potential solutions
Evolving software and new ECUs require hardware-independent, cost effective software solutions that deploy to various SoCs. Emerging industry solutions will likely involve the use of one of these embedded operating systems:

Linux was introduced more than 30 years ago and is an open-source (cost-free) operating system. It powers most of the internet in addition to mobile and IoT devices, cloud computing platforms, desktop computers, and other technologies. Linux is scalable and customizable, enabling seamless integration of future innovations and cross-pollination of software technologies. With RedHat software for Linux OS recently gaining functional safety certification, the system is mature and poised for automotive software applications capable of becoming hardware-independent. Despite its many benefits, there are challenges associated with working with Linux OS, which include working with the open-source community to implement new features, resolving software defects and complying with various open-source licensing agreements.              

QNX was developed by Blackberry and introduced in 1982. QNX is today’s most practical solution for the high-performance SoC embedded systems market. QNX Hypervisor helps address complexities related to safety and cybersecurity to ensure freedom from the interface and manage security vulnerabilities. With functional safety certification and a wide variety of library support (for example networking, audio, graphics, ADAS platform), QNX has established itself as a market leader in automotive applications. According to QNX, nearly a quarter-billion vehicles run QNX software.

Other commercial embedded OS suppliers include Green Hills Integrity OS, Wind River VXWORKS, and SYSGO PikeOS.

Emerging industry standards
As capabilities and markets for ADAS expand, government agencies and other global organizations have attempted to keep pace by regulating the market and setting a uniform standard for modern ADAS capabilities. These organizations include the National Highway Traffic Safety Administration (NHTSA) and Insurance Institute for Highway Safety (IIHS) in the USA, the European New Car Assessment Programme (Euro NCAP) and the Society of Automotive Engineers (SAE), which has global reach.

On the software side, the AutoSar Adaptive platform offers a standardized definition of software for future high-performance computing ECUs. The Adaptive platform differs from AutoSar Classic in that the latter, which has been around for more than 20 years, offers static and standardized software development in ECUs.

Launched in 2017 to address those questions for which Classic AutoSar Standard was not intended, Adaptive AutoSar uses a more flexible approach, enabling the dynamic deployment of new services. Additionally, Adaptive provides the interfaces necessary for integration with third-party software via non-AutoSar systems. AutoSar Classic retains its usability for low-performance, high-reliability, high-safety ratings (ASIL-D) and real-time performance needs. While AutoSar Adaptive provides a solution for high-performance applications that require additional computing power and flexibility, its slower-paced standards development and potentially prohibitively expensive license fees make the solution less attractive, leading to slower adoption.

The challenges of a cost-competitive space with a history and culture of unique hardware and software loom large, although emerging industry-standard solutions can help level the playing field. From a manufacturer’s standpoint, the greatest needs are creating more cleverly managed upfront investments, continuing development of hardware-independent software platforms and avoiding over reliance on unique hardware accelerators while attempting to migrate most, if not all, software functionality into hypervisor-based software architectures. The goal is to prevent over-reliance on unique hardware accelerators, which aligns with SoC’s aim to migrate almost all software functionality into hypervisor-based software architectures. The result will be architectures that quickly port and deploy across various hardware platforms and at lower development and validation costs.

link

Leave a Reply

Your email address will not be published. Required fields are marked *