AWS for Industries

AWS and Qualcomm Software-Defined Vehicle demonstrator for Cloud-Native Snapdragon Digital Chassis development

With vehicles becoming increasingly intelligent and connected, software is playing an increasingly important role in their design and evolution, accelerating the path to the Software-Defined-Vehicle (SDV). As SDV gains traction, the emergence of Cloud-Native development for automotive applications is becoming pivotal. In this blog we will discuss the latest breakthroughs and trends for SDV and showcase how automakers can use AWS and Qualcomm technologies to scale and accelerate vehicle software development. Using the Snapdragon Digital Chassis enabled cloud-native development, powered by AWS, automakers can access the capabilities needed to develop in the cloud, test, and validate code on virtualized digital twins of the vehicle hardware that it will eventually run on.

Continuous Innovation at Qualcomm and AWS

Last year at CES 2023, Qualcomm announced Snapdragon Ride Flex SoC – an in-vehicle high-performance, central-compute platform capable of hosting and executing flexible mixed critical cloud-native workloads.

In addition, Amazon Web Services (AWS) and Qualcomm Technologies, Inc. announced a long-term commitment to provide a comprehensive development infrastructure and tools aimed at helping the automotive industry’s transition to being software-defined. Most recently, AWS announced the general availability of Amazon Elastic Compute Cloud (Amazon EC2) DL2q instances, powered by Qualcomm® Cloud AI 100 for cloud AI inferencing.

At CES 2024, the companies joined efforts again to demonstrate the potential of a Snapdragon Digital Chassis enabled cloud-native development, powered by AWS. This demonstration is designed to showcase how automotive application development and deployment expedite the evolution of software development practices and increases developer efficiency.

These technology advancements incorporate mature infrastructure and modern software development workflows, enabling agile development and deployment of new features throughout the vehicle’s lifespan.

Overcoming Barriers to entry for SDV

To enable software defined vehicles, automakers need to create software collaboration infrastructure that enables software from its internal stakeholders and external partners to be integrated and delivered into the car in an agile way. This creates significant integration, verification and validation challenges considering safety requirements of in-car software. It also introduces additional requirements for Automakers in terms of development process and resources which can impact the timeline to implement these new collaborative software workflows.

Advancements In Cloud-Native Automotive Software Development

To overcome these challenges of adopting modern software development workflows, automakers are now creating “Automotive Software Factories” (ASF) using the latest DevOps methodologies for software development, like the ones described in this repository and blog post.

Cloud-Native OEM Automotive Software Factory workflow. We start with what is running in the cloud. A first pipeline refers to Developers. Developers develop, review code, perform unit testing, static analysis, and then an artifact is automatically built. A second pipeline refers to data ops: data is ingested from vehicles and make available for reprocessing and testing tasks. Dad is used to test the artifact that has been built and deployed in the virtual SoC running in the cloud. When the testing results are satisfactory, the artifact is deployed on HiL and real vehicles to continue the testing.
Figure 1: Cloud-Native OEM Automotive Software Factory: Enabled by Qualcomm’s Car to Vehicle and AWS Infrastructure

The main goal of an ASF is to offer a consistent centralized software development environment where teams from the OEMs, suppliers, and partners can access and collaborate together.  Adding the ability to test and validate the software on a virtual representation of the physical hardware (commonly referred to as Digital Twins or Virtual Electronic Control Units, vECUs), with functional parity is very powerful. With these new capabilities, automakers have a scalable solution needed to accelerate vehicle software development. Digital Twins and vECUs form a crucial infrastructure component of ASF.

The Solution: Expanding the role of “Software-in-the-Loop” (SiL)

Within the Snapdragon Digital Chassis product portfolio, Qualcomm Technologies has enabled cloud-native development infrastructure technologies, including Snapdragon Virtual Platforms. These platforms provide a virtual representation of the physical SoCs used in vehicle ECUs. As part of this latest collaboration, AWS and Qualcomm are enabling Snapdragon Virtual Platforms on AWS, using AWS Graviton processors and Amazon DL2q instances powered by Qualcomm® Cloud AI 100. Altogether, this demonstrator aims to showcase a high-performance emulation solution with the ability to run the same base platform software that will run on the actual silicon within the vehicle’s ECU hardware, providing environmental parity between the cloud development infrastructure and the physical ECU that the software will eventually run on.

This ASF infrastructure reduces dependency on the physical vehicle hardware in the development phase: developers can work anywhere in the world without the need to have a physical evaluation kit or samples of the real hardware unit on their desks. Results from a white paper by Marelli in developing software in an environment that leverages a Digital Twin for an infotainment electronic control unit indicates a 70% reduction in time for developing new features, and 30% reduction in prototyping costs. Furthermore, incremental testing can be executed in SiL on digital twins of the real hardware, providing additional “shift-left” in development operations. Last, this setup has the potential to significantly boost testing coverage: current practices are highly dependent on HiL farms, which are limited in resources and come late in the process. The solution presented executes tests with high levels of parity leveraging cloud scalability as soon as the very first versions of the software are available. Following these premises, developers integrate software into a shared cloud platform and test it against the entire software setup, following a DevOps workflow essential for SDVs.

Solution Architecture: Demonstration of Cloud-Native Automotive Software Development

The demonstration features a cloud-based developer environment, as well as virtualized Snapdragon SoC platforms to test and validate software in the cloud. The developer environment can be used as a central point for the development of both advanced driver assistance systems (ADAS) and infotainment software that run on the Snapdragon Digital Chassis platform. The architecture is visualized in figure 2 below:

In AWS cloud it's presented a pipeline for developers to write and build code. This code is then tested in a digital twin of the real hardware, Qualcomm's SA8775. In the digital twin, we have an hypervisor, Gunyah, RedHat In-Vehicle OS, AAOS, Podman/BlueChi and a set of microservices that implement an ADAS functionality. This runs on G5g.metal> Other instances are used to run an IDE provided by Qualcomm's Code Collaboration environment, A portal to manage Car2Cloud and Services, and a backend to run inference, leveraging EC2 DL2q. When the software is ready for, it will be deployed on a physical Qualcomm's SA8775 that will run the same exact software stack Figure 2: Architecture of CES 2024 Demonstration of Cloud-Native Automotive Development for Snapdragon Digital Chassis

Figure 2 showcases the technology demo setup featuring two key elements:

  • SA8775 Physical HPC (ADAS & Cockpit): The physical deployment infrastructure utilizing Snapdragon Ride Flex SoC based on the Snapdragon
  • SA8775 Virtual HPC (ADAS & Cockpit): The automotive development infrastructure hosted on AWS EC2 instances which comprises of Qualcomm Code Collab (QCC) (a web-based Integrated Development Environment provided by Qualcomm) and Snapdragon Ride Flex SoC-based Virtual Platforms. This runs the same configuration as the physical platform providing functional environmental parity between virtual and physical platforms.

An Automatic Emergency Braking application is used as an illustrative workload within the demo, implemented as microservices within containers. Both the physical and virtual setups utilize a pre-recorded simulated video as the input stream for testing purposes. The flow shown in figure 2 is described as follows:

  1. Default setup has a simulated video stream being outputted from Flex SoC based ADP on a monitor with bounding boxes across certain objects detected.
  2. As part of development phase, developer logs into QCC and clones the AEB GitHub project. The developer then makes a modification in the code to enable detection of additional objects and then triggers a build.
  3. As part of the testing phase, the generated application binary is then executed on Flex SoC based virtual platform which is running the same software configuration as the physical platform. On successful completion of the testing phase, the tested application binary is transferred to Qualcomm’s C2C Connected Services Dev Portal where user can subscribe to the new functionality provided by the new AEB binary.
  4. Post subscription, the new AEB application binary gets deployed on Flex SoC ADP with binary parity and the new updated video stream should start showing additional object detections with bounding boxes like what was tested in the cloud.

Conclusion

In conclusion, this blog has highlighted the collaborative efforts of Qualcomm and AWS in developing technologies that enable efficient automotive software development. Leveraging Snapdragon SoCs and Qualcomm Software Platforms on AWS enabled cloud infrastructure, these innovations enhance the overall efficiencies of OEM’s automotive software development resulting in a faster time to market of bringing in new user experience to end customer.

To learn more about this solution, please visit the AWS or Qualcomm in West Hall at CES 2024:

  • AWS booth #3901
  • Qualcomm booth #4901

The demo will be made available as a video after the event.

TAGS:
Stefano Marzani

Stefano Marzani

Stefano is focused on helping to solve the biggest challenges in the automotive industry. His current focus is helping the automotive industry transition to software-defined vehicles, enabling autonomous functionalities, mobility fleet solutions, and delightful user experiences. Stefano’s technical expertise lies in IoT, Machine Learning, Vehicle Architecture, HMI, and Automotive Software Development & Tooling.

Girish Shirasat

Girish Shirasat

Girish is a Senior Director of Product Management at Qualcomm responsible for product definition and system software architecture for Qualcomm’s next-generation Software Defined Vehicle (SDV) enabling solutions. Along with defining Qualcomm’s automotive SDV related products, Girish also focusses on working with automaker to help build their next-generation car to cloud SDV solutions based on Qualcomm’s product portfolios. Girish also represents Qualcomm in consortia like SOAFEE and Eclipse SDV.