This blog is part one in a three part series. Look for part two next week.
Automotive Grade Linux (AGL) is one of the most innovative community involved in embedded systems developments. Such innovation, as officially recognized by Embedded Computing Design, has its root in the open source and code-first approach and consists of delivering a de-facto Linux-based standard platform that significantly reduces automotive software time to market and licensing costs. This platform is a disruptive innovation in the automotive industry today, where OEMs and Tier-1 companies are continuously challenged to provide smarter, cheaper, and more connected functionalities.
The AGL software platform provides a vertical solution starting from the operating system layer (Linux) up to APIs, software-development kit (SDK), and reference applications. However, autonomous driving and software defined vehicles require the execution of several hundreds functions with different functional safety requirements. In fact, users want to see the automatic pilot function running together with their mailbox application, or to be able to let children play the latest 3D video games on the backseats while using the rearview camera. In this context, there is a need to extend the AGL platform to support an additional technology allowing the safe and efficient combination of mixed critical functions: virtualization.
What is virtualization?
Virtualization is a technique used to create multiple virtual execution environments by means of resource abstraction or partitioning. Historically designed as a technology for the server market, with the increasing computational power of embedded hardware platforms, virtualization also found applications in embedded systems with hypervisors, system partitioners, and containers. The hypervisor is a software component running close to the operating system layer that creates hardware resources abstraction for the execution of different operating systems concurrently in the form of virtual machines (VMs). System partitioners do not use abstraction, but partition system resources to create siloed virtual environments for the execution of operating systems directly on the (partitioned) hardware.
Containers, on the other hand, achieve the same objective of hypervisors and system partitioners, but without using specific capabilities of the hardware platform. Their use in AGL is limited to workloads without safety or real time requirements for this reason.
AGL automotive virtualization
In modern vehicles there is a need to isolate the different functions securely, with support for ISO 26262 safety certification, and in a way that they can’t interfere with each other. Moreover, users are expecting an increasing number of these functions to be installed on the fly, customizable and upgradable over the air. In this scenario, adding a new electronic control unit (ECU) for each new function in the car architecture does not apply anymore: it does not scale with the increasing number of functions, it cannot provide the customization level required by the users, and it exponentially increases costs (hardware, wiring, maintenance, etc.).
For AGL, virtualization is the software defined connected car enabler technology because of its ability to host the execution of several (and isolated) functions concurrently, by offering at the same time run-time configurability, a flexible and scalable architecture in a secure and efficient way. AGL is working to expand its platform with a set of virtualization technologies and, thanks to a team of virtualization professionals inside the community that constitutes the Virtualization Expert Group (EG-VIRT), published the white paper, The AGL Software Defined Car Architecture.
The objective of this work is to pave the way towards an open source de-facto standard for software defined connected cars.