In automotive, high-performance compute (HPC) supports the increasingly demanding processing requirements of the software-defined vehicle, thereby enabling a feature-rich platform that can endure for a vehicle’s lifetime.
As OEMs seek to centralize compute, HPC platforms offer a low-latency solution to process data from cameras, radars, wireless networks and more — delivering the seamless advanced driver-assistance systems (ADAS), infotainment, user experience and connectivity features that consumers expect.
Once a vehicle is on the road, these hardware platforms must be robust and updatable to keep up with the ever-increasing demands of software — particularly as vehicles move toward higher levels of autonomy.
Reducing complexity with HPC
Distributed approaches to vehicle architecture are becoming increasingly unsustainable as consumers demand more intelligent features from today’s vehicles. Relying on a dedicated electronic control unit for each feature complicates development and validation, and uses too much physical space. As a result, OEMs are up-integrating features into centralized HPC platforms as part of an overall shift to a zonal architecture, which helps reduce wiring, weight, costs and complexity.
Rethinking vehicle software architecture
The software architecture is critical to enabling the HPC platform to evolve its capabilities over time. With an architecture that decouples hardware from software, HPC hardware can be replaced as needed. Decoupling hardware and software also means that software development, deployment and upgrades can occur on their own schedules, without being tied to the limitations of the hardware. This is particularly important in the automotive industry, where multiple third parties provide software for OEMs.
The approach requires a vehicle software architecture with several key ingredients:
- Application programming interfaces (APIs): APIs enable abstraction and interface standardization by providing a path for lower-level devices such as sensors and actuators to present a standardized set of inputs to higher-level software.
- Service-oriented architecture (SOA): SOA, also known as a service-based architecture, is a development approach that enables the reuse of software by creating independently deployable software components called services. Services perform tasks automatically, react to events or fulfill data requests from other software. Each service serves a specific function and can communicate with other relevant services across operating systems and programming languages.
- Software containerization: Software containerization provides the flexibility OEMs need to modify individual functions rather than issue updates to entire monolithic code bases — while ensuring that any changes will not negatively affect adjacent safety-critical software in the vehicle.
- Hypervisor: Hypervisors enable HPC platforms to dynamically allocate compute resources in real time in response to changing road conditions. For example, a vehicle driving along an empty road might have plenty of compute power to run infotainment and in-cabin user experience systems at full capacity, but when the vehicle approaches a construction zone — with signage, lane changes and heavy traffic — the HPC platform can shift computing resources to process the high volumes of new sensor data.
Withstanding automotive environments
In IT and cloud applications, HPC platforms most often operate in climate-controlled and carefully monitored environments. IT personnel have easy access to the hardware, enabling full lifecycle management and making it easier to continuously update the hardware to address ever-increasing demands for more processing power to run advanced applications.
However, in automotive environments, HPC platforms must be able to perform under the harshest conditions, which requires high-performance systems that are robust enough to withstand temperature extremes and high-vibration forces.
Aptiv’s high-performance compute solutions
As the only provider of both the brain and nervous system of the vehicle, Aptiv delivers the foundational hardware and software elements that enable the software-defined vehicle, including flexible high-performance compute solutions.
Aptiv’s central vehicle controller (CVC) is an HPC platform that translates software code into physical action. It can be a power and body controller, propulsion and chassis controller, data network router, gateway, firewall, zone master and data storage hub all rolled into one — or it can perform a mix of some of those functions.
However, to enable the vehicle to meet evolving safety regulations and support higher-level functions, newer hardware could also become necessary. The right software and hardware architecture must allow physical upgrades to just the specific compute components that need it the most, well into the future.
That’s where Aptiv’s open server platform (OSP) comes in. The OSP provides an exchangeable high-performance compute solution to enable the vehicle to truly evolve over the vehicle’s lifetime. OSP modules can be designed to stand alone or be integrated with a central vehicle controller, optimizing costs and making OSPs easy to swap out when an upgrade is needed.