Home > AI Solutions > Artificial Intelligence > White Papers > Dell Automotive Reference Architecture > When two worlds converge
For decades, the automotive industry’s primary emphasis has been on manufacturing efficiency and is focused on incremental improvements and enhancements with each generation of vehicle up to the present time. In contrast, the IT industry has played a pivotal role in the evolution and advancement of digital technologies, encompassing software and hardware development, compute, IT services, storage technologies, and networking connectivity. The emergence of autonomous driving and electrification has paved the way for novel markets and the inception of automotive startups implementing new technologies and in-vehicle architectures. These new entrants, distinct from traditional automotive backgrounds, originated in the realm of programming and computer hardware and software, are forcing a change in the automotive industry. A paradigm shift is occurring with vehicle manufacturers increasingly recruiting computer programmers and architects rather than solely focusing on traditional design and manufacturing practices. This trend echoes the convergence of the Telco and IT industries.
The following figure shows a VENN diagram that illustrates what happens when two disparate industries begin to overlap:
In this case, the IT and automotive industries each have their strengths and weaknesses. Each industry has grown up in isolation from each other until now. AD and electrification have allowed startups and now mainstream car manufacturers to rethink how the traditional car is architected and manufactured in an increasingly electric and autonomous vehicle landscape.
One such overlap that is beginning to form between the automotive and IT industries is with V/V of autonomous vehicles. Currently, HiL Rigs serve as the standard method for testing and validating the ECU’s performance in vehicles, acting as a crucial link between IT and automotive domains. A traditional HiL Rig facilitates communication between a computer and an ECU by converting the unique communication protocols employed by the vehicle into Ethernet signaling. The challenge with HiL Rigs is that they are prohibitively expensive, difficult to scale, proprietary, and require engineering domain knowledge.
A new type of HiL Rig, a server-based HiL Rig, bypasses the unique automotive interfaces and communicates directly to the ECU controller memory through a PCIe bus. The significance of the PCIe link cannot be understated. It is the bridge between two domains that allow IT best practices to validate and verify ECU AI model performance. ECUs can be built on PCIe plug-in cards and integrated into servers along with GPUs to accelerate the SiL process by introducing real hardware sooner in the AI development pipeline and without the aid of a traditional HiL Rig. This new architecture is not only scalable, but less expensive and can significantly reduce development cycle times by dramatically accelerating the SiL process by communicating directly to hardware. Are traditional HiL Rigs still required? The short answer is yes. The process of integrating and testing all sensor interactions with the ECU is still required, but not at the scale previously required.