Qualcomm Snapdragon SoC Marries Infotainment and ADAS Support

In many respects, the automotive sector may be the field undergoing the most innovation and technological change in the entire industry. With the simultaneous rise of electric vehicles (EVs), advanced driver-assistance systems (ADAS), and autonomous vehicles (AVs), today’s technologies are rendering the modern automobile almost indistinguishable from those 10 years ago.

Supporting this changing landscape will require new concepts in hardware and software which increase flexibility, performance, and cost-efficiency. To help enable this future, Qualcomm announced a new automotive SoC family at CES last week.

In this article, we’ll discuss some technological shifts in the automotive industry, including the transition to a zonal architecture and how Qualcomm’s new SoC hopes to enable these changes.

 

Ride Flex SoC

To address the new concepts in the changing automotive landscape, Qualcomm has announced the release of their new Ride Flex automotive SoC family.
 

The Snapdragon Ride Flex SoC. Image used courtesy of Qualcomm

 

As the SoC is expected to begin production in early 2024, not many technical specifications are currently known. What we do know, however, is that Qualcomm is claiming that the Flex SoC is designed specifically to support mixed-criticality workloads across heterogeneous computing resources on a single SoC. What this means is that the SoC will be able to simultaneously support digital cockpit (in other words: infotainment), ADAS, and AD functions on one chip.

 

From Domain to Zonal

From a hardware architecture perspective, one of the major shifts happening in the modern automobile is a change from a domain architecture to a zonal architecture.

When electronic control units (ECUs) first started getting integrated into vehicles, there were relatively few functions that relied on electronic control. Because of this, it made sense for automotive designers to incorporate a single ECU for each specific function.

Here, ECUs that shared a common function would be grouped together under a single gateway or domain controller. This architecture, where in-vehicle electronic systems were grouped by function, is known as the domain architecture. 

 

Domain versus zonal architecture. Image used courtesy of Texas Instruments (Click image to enlarge)

 

However, as the rise of ADAS and AVs has caused the modern vehicle to become more densely integrated with sensors and added functionality, the domain architecture has become infeasible. Specifically, the cost, weight, and complexity of the electrical wiring that interconnects each domain is becoming a limiting factor as more domains are added.

Instead, the industry is seeing the rise of zonal architecture to resolve this issue. The zonal architecture refers to a new automotive architecture in which the ECUs within a vehicle are grouped by physical proximity into a number of unique zones. Within this, each zone has a single zonal controller which is then connected to a central computing cluster in the center of the vehicle.

Because of the prioritization of spatial locality, the zonal architecture results in an intelligent vehicle that has reduced cabling, leading to less cost, complexity, and weight.

 

Needs of a Zonal Architecture

While the zonal architecture is paving the way for intelligent vehicles with more added functionality, implementing this architecture still requires some technological advances of its own.

 

Examples of varied sensors within a modern vehicle. Image used courtesy of Guerrero-Ibáñez and co-authors

 

One major challenge with the zonal architecture is designing a central compute cluster that is flexible enough to handle the varied workloads within a vehicle. As evidenced by the need for a zonal architecture, modern vehicles consist of a plethora of different workloads, each with its own unique computing requirements.

Consider, for example, that an intelligent vehicle may need to simultaneously support functionality such as graphics processing for infotainment systems, sensor fusion amongst the vehicle’s varied sensors, and AI/ML inference (in other words., computer vision) for ADAS and AV functionality.

 

Central Compute Clusters

As a result, the central computing cluster of a zonal architecture has the unique requirement of being versatile enough to handle each disparate task while also being performant enough on each task to support the real-time decision-making required by ADAS and AV.

With all that in mind, Qualcomm is hoping to pave the way for the zonal architecture by providing flexible hardware to support central compute clusters.