Co-authored by James Scobie and Govind Wathan
The automotive industry is rapidly transforming, with technology driving innovation, automotive standards shaping requirements and consumer preferences changing demands. These factors are significantly impacting all automotive applications from Advanced Driver-Assistance Systems (ADAS), autonomous drive, In-Vehicle Infotainment (IVI) and digital cockpit, to powertrain and chassis. Arm offers a broad range of different classes of processor, specifically designed to address the needs for each of these automotive applications.
To help you take advantage of this processing power, here is a guide to how Arm’s different classes of processors can be applied to various automotive applications.
There are typically three types of ADAS systems in an autonomous vehicle, which are camera, LIDAR and radar systems. Depending on its type and class, these ADAS systems can be powered by a combination of Cortex-A, Cortex-R and Cortex-M processors. Arm sees the application of processing power fall in the following four stages for autonomous processing:
Sense > Perceive > Decide > Actuate.
This stage involves gathering information about the environment through a range of sensors. Depending on the type of sensor, the data collection systems within the vehicle can be powered by Cortex-A, Cortex-R or Cortex-M processors. For example, in a forward facing vision system, the sense part is likely to use an ISP (Image Signal Processor) and a Computer Vision accelerator combined with a Cortex-A processor, and possibly a Cortex-R or Cortex-M. Radar is another sensory system which can be addressed with different processing capabilities. Generally, radar and LIDAR are supported well by Arm Cortex-R processors but can also be supported by Cortex-A processors such as Cortex-A55 with a Cortex-M. Simpler sensors points throughout the vehicle all combine to provide a wider picture of the state of the vehicle and its environment - detecting and measuring the vehicles actuation, speed and condition as well as inputs from things such as ultrasonic sensors, which are used in parking and close range maneuvering. These are generally a Cortex-M based application.
In the perceive stage, the system is filtering and interpreting the data, understanding what can be learned about the environment from the sensor data. The types of workloads seen in this stage are throughput sensitive which is why the Cortex-A processors play a predominant role here. A lot of data in ADAS and Autonomous systems is parallelizable, that is, it can be processed at the same time. For this reason, Cortex-A processors are usually closely coupled with GPUs and ML processors, thus enabling a large amount of computation of the sensor data.
During the decide stage, the system safely chooses the correct action for the vehicle to take. There are increased levels of performance, latency sensitivity and safety required. Cortex-A plays a critical role as a high-performance decision and planning engine. Its superior ability in cognitive processing through Arm DynamIQ technology makes it a perfect fit for autonomous vehicles, where more powerful machines are needed. Cortex-R also comes into play here, particularly from a functional safety perspective. It can act as a sense check on the decision made by higher level processing power before reaching the actuation stage.The Cortex-R52 is a good example of this due to its focus on redundancy and determinism.
The actuate stage is where the vehicles dynamic behaviour is modified by the decision reached in the earlier processing stages. At this point, messages are distributed across the vehicle that require deterministic behaviour and real-time responsiveness. Cortex-R is a natural fit for this step in the processing pipeline particularly where there is demand for highly intelligent control and communication. Depending on the complexity of the control tasks and the actuation demands, Cortex-M processors can also handle some specific and simple actuate tasks. Functional safety requirements in this stage are rapidly growing from ASIL B which is causing a migration from the use of Cortex-M processors to Cortex-R processors at this stage.To achieve Level 3 autonomous drive, ASIL B requirements must be met by the system, whilst for levels 4-5 autonomous drive, ASIL D requirements need to be met by the system particularly in the higher compute point in the process. The functional safety Standard ISO 26262 defines different levels of functional safety in terms of ASIL with the highest level being ASIL D. Functional safety is an important requirement for autonomous applications with increasingly higher demands on the ASIL moving from ASIL B and ASIL D as the autonomy increases. Arm understands the importance of building functional safety into designs right from the start – read our recent functional safety blog to find out more on how Arm is enabling functional safety technology and the solutions we offer.
IVI/ digital cockpit is one of the largest growing markets for vehicles and is seeing significant innovation advancements. IVI systems were historically the radio and media player but have advanced in recent years to include climate control, navigation, media streaming, gesture recognition, gaming and Heads-Up Display (HUD). The total number of touchscreen displays in a vehicle has also gone up, extending the IVI user experience to passengers and all-in-all, increasing the demands for performance.
Infotainment systems have been influenced by the technology and experiences from the smartphone ecosystem. Cortex-A processors can satisfy the processing requirements as well as the safety requirements that come with the move towards digital cockpit. Arm big.LITTLE technology is widely used in the IVI space, enabling a further tuned balance in performance and power efficiency. The recently announced Cortex-A76 is the latest high-performance CPU to support big.LITTLE when paired with the high efficiency Cortex-A55.
The next generation of IVI systems are merging with cluster systems, which display vehicle and driver information to form a more immersive experience in the vehicle. This combined system is referred to as the digital cockpit. The inclusion of the cluster which displays instrumentation and control for the vehicle's operation, mean that digital cockpit systems have mixed-criticality from a functional safety perspective. Moreover, these systems also demand increasing levels of performance. Cortex-A CPUs such as the Cortex-A76 deliver a 20% uplift in performance over its predecessor and is designed on a rigorous design flow to avoid systematic faults.
This merging of what have been traditionally separate systems brings interesting changes and challenges for both hardware and software developers. The software stack now needs to be able to run safety-critical and commercial applications on the same System-on-Chip (SoC) and display these different applications safely on their respective displays. The latest Cortex-A processors provide the performance required to run a Rich OS next to a RTOS or Autosar stack, and the Cortex-A offers hardware support for the hypervisor platform that enables it.
Today, there are two approaches to IVI systems.
Previously we have seen a small amount of functional safety required for the cluster systems, where tell-tales inform the driver about speed and hazards. This level of functional safety has been typically more ASIL B than ASIL D.
Due to the growing demand for cleaner and more efficient vehicles, the largest growth area for powertrain is in its electrification. Conventional Internal Combustion Engines (ICE) will continue in production for years to come with yet tougher, demanding emission controls, but increasingly in hybrid configurations. Full electrification adoption will remain dependent on energy storage technology and infrastructure for charging points and energy distribution, but over time we’ll see the balance move to electric drives with higher power battery management.
There is high demand for both ICE and electrification and so solutions to meet these needs are essential. Arm can address the needs for both ICE and electric drives.
Functional safety is a critical part in drivertrains, a requirement which can be addressed through the use of a processor with systematic capability for ASIL D, such as the Cortex-R52.
Applications of body electronics are ubiquitous, covering the entire vehicle. Body applications range from simple, small sense and actuation nodes, like door locks, temperature and position sensors in Heating Ventilation Air Conditioning (HVAC) control, to more complex central Body Control Unit applications, sometimes incorporating vehicle network control. With a range of different Tier 1 suppliers going into many different vehicles, being able to build multiple applications onto a single platform based on a common architecture gives the flexibility, scalability and improved reuse of both tools and software. Arm’s Cortex-M class of processors delivers a simple programming model and well supported ecosystem of tools and software, together with highly configurable options to tailor the hardware. This reduces development cost and time to market whilst meeting the demanding needs of these systems.Many of these applications are key-off, meaning they need to remain operational when the car is parked. Low energy consumption is important both at run-time and during these "key off" periods to maintain battery life. Cortex-M processors enable body applications to wake up, complete an action and go back to sleep rapidly, with the ultra-low energy sipping power normally associated with wearable applications. Cortex-M processors easily meet these needs, and also benefit from a supportive ecosystem, tools and software that allow easy implementation which improves efficiency and reduces development cost. Security is increasingly important across a broad range of automotive applications and this includes those in body control. The introduction of Arm TrustZone within the Cortex-M family offers an opportunity for added security support in this profile of processors for these applications. Cortex-R processors also offer a flexible solution for high performance centralised body control systems and network controllers. The desire to keep a check on the increasing number of ECUs in the vehicle makes Cortex-R processors an ideal fit where consolidation of functions into fewer ECUs is needed by simplifying the migration of multiple applications into a single processor and maintaining their isolation.
Arm provides total automotive compute capability for every application within the vehicle helping to meet increasing demands in innovative new ways, even in this incredibly fast-moving and changing industry. If you would like to find out more or have any questions on how Arm processing power can be applied to your automotive applications, please watch our on-demand automotive webinar below.