Supercharging Development for Automotive E-Cockpit on AWS: Achieving Bit Parity

As automakers strive to enhance the in-cabin user experience, automotive development teams are tasked with delivering more features and integrations than ever before while balancing trade-offs between complexity and maintenance. Supercharging development for an automotive e-cockpit on Amazon Web Services (AWS) helps development teams to accelerate software output while achieving bit parity. With the advent of the Software-Defined Vehicle (SDV), automakers are shifting to software-centric practices to enable capabilities for continuous software delivery, and move toward “software-first” strategies throughout the vehicle lifecycle.

In this blog, we explain the trends and challenges associated with improving vehicle software code quality and the vehicle development process. You will learn how you and your organization can overcome hardware constraints while helping developers to take a software-first approach without sacrificing performance and responsiveness. We will provide details on how automakers can use AWS and Panasonic Automotive technology to help achieve hardware analogous performance using Panasonic Automotive’s vSkipGen—a product that helps software teams to develop target code without the hardware, further optimizing the product-development process and improving product quality.

A transition to “software-first”

As the automotive industry undergoes a significant transformation towards a “Software-First” approach, characterized by a shift to software-defined architectures, developers are compelled to embrace rapid innovation.

This shift necessitates the adoption of Continuous Integration/Continuous Deployment/Continuous Testing (CI/CD/CT) DevOps methodologies to facilitate frequent software updates. These updates, integral during both development and post-launch phases, offer a spectrum of benefits ranging from quality enhancements to the introduction of new features. It is imperative to note that within this dynamic landscape, systems requiring safety certification, which preclude the utilization of CI/CD/CT practices, must be carefully isolated. This isolation ensures that the integrity of safety-certified systems remains intact, safeguarding against potential risks while allowing for the advancement of non-certified software components.

The quality of a software update is strongly correlated with the automaker’s adoption of a software-first approach. In other words, weaker levels of adoption tend to lead to quality issues that overtake the development team’s time. That impact is then carried over from vehicle development to launch. These quality issues can cause significant budget overruns, diverting resources from other key initiatives. Both automakers and customers can experience delays, and automakers are sometimes unable to deliver new features.

The Panasonic SkipGen evolution

Panasonic Automotive Systems Company of America (Panasonic Automotive), recognized as a leader for cockpit reference platforms, has decided to advance the positioning of their solution by offering a virtual replica of the hardware reference platform SkipGen 3.0. This solution, officially named Panasonic Automotive vSkipGen (short for Virtual SkipGen) targets automotive developers early in the vehicle development cycle. Panasonic Automotive vSkipGen enables teams to increase software output while maintaining bit parity, a definition used to describe a software compilation for a virtual target that can also run in binary parity on its physical counterpart. By leveraging industry standard Virtual Input / Output (VirtIO) compliant device abstraction layer, vSkipGen offers a number of guest operating systems such as Android Automotive (AAOS) and Automotive Grade Linux (AGL). The use of virtual devices in the early stages of automotive software development is referred to as “shifting left.” Shifting left describes the relationship between the automotive SPICE V-model development lifecycle phases and helps developers to perform activities in a parallel and iterative manner as opposed to the traditional serialized workflow, where software development tends to fall to the right of early hardware development.

Hardware In-The-Loop Scarcity

Traditionally, software development teams had to wait for early samples of in-vehicle infotainment (IVI) and digital instrument cluster (IC) hardware to start their software development, integration and testing activities. Due to limited production runs, these early hardware parts are typically higher in cost, less likely to have stable firmware, and require construction of a physical workbench with compatible wiring and power in order to begin application development on this hardware and integration testing between vehicle systems. This limited hardware availability, usually coupled with global development teams who need access, creates logistical challenges around resourcing hardware build and work group priority. As a result, the traditional workflow is characterized by strict adherence to the V-model in order to save costs, reduce risks, and stabilize core firmware prior to further development. The overall result is that early hardware-in-the-loop (HIL) test setups are expensive and timely to procure, setup, maintain, and do not scale well across teams and geographies. Scarcity of HIL test environments needed for software tests reduces developer productivity, and can delay time-to-market of the vehicle.

Hardware dependency risks

To illustrate the hardware dependency risks for Automotive development teams, consider a use case in which the developer of an IVI application needs data from the IC application owned by another team. The IVI application developer needs to wait for the physical IC as well as compatible firmware to be able to begin testing. The flashing of firmware onto devices, if not done correctly, introduces risks, including disrupting the hardware’s ability to boot—known as “bricking” the target. Bricking adds downtime, because not every team member may have the skill set or tools to recover the hardware to a usable state. Firmware versioning across multiple hardware components within HIL setups also introduces firmware compatibility challenges, which can cause blockers for development and validation teams.

The solution

On November 27, 2023, Panasonic Automotive announced the release of vSkipGen on AWS Marketplace, which is a place to find, test, buy, and deploy software that runs on AWS. vSkipGen comes with fully optimized support for Android Automotive IVI, AGL, and any POSIX-compliant operating system. vSkipGen helps automakers to decouple the software development from the hardware, reduce time to market, and improve software quality. This improvement includes providing automotive development teams with access to virtual devices, like audio, location (GNSS), graphics, vehicle networks, and more, using OASIS standardized interfaces of virtIO.

vSkipGen provides access to hardware-accelerated workloads, such as multimedia processing, and renders high-resolution user interfaces for responsiveness by using AWS Graviton processors. AWS Graviton-based instances are designed to deliver up to 40% better price performance over comparable x86-based instances, while providing secure and resizable compute capacity for virtually any workload. With processor options such as t4g GPU (NVIDIA), automotive developers can develop and test high performance computing (HPC) applications and toolchains using the same software that is also targeted in embedded automotive platforms. This capability helps to directly run the ARM Instruction Set Architecture (ISA) without the need to cross-compile.

The opportunity

Panasonic Automotive designed vSkipGen on AWS to deliver an advanced spectrum of driver assistance and IVI features available to developers, by click of a button. This combination offering is designed to the needs of automakers and suppliers developing IVI and IC applications virtualized on AWS Graviton processors.

Furthermore, vSkipGen opens the opportunity for automotive teams to accelerate their deliverables by simulating dependencies of their IVI and IC components and performing higher-level integration testing. These capabilities help to prevent delayed defect detection, saving time and money. For such simulations, developers can use common vehicle communication protocols such as Scalable service-oriented middleware over IP (SOME/IP), Controller Area Network (CAN), Data Distribution Service (DDS), and Message Queuing Telemetry Transport (MQTT) to communicate with virtual embedded development targets (vEDT) in a simulated vehicle electric/electronic (E/E) network. A vEDT is meant to characterize different group functionalities, known as Electronic Control Units (ECUs), within the vehicle.

Solution architecture

Figure 1 below illustrates how vSkipGen uses AWS Graviton-based instances to run the cockpit domains, consisting of AGL and/or AAOS, as virtual machines in the cloud. Panasonic Automotive built vSkipGen for virtualization along with the necessary peripherals on the Amazon EC2 G5g instance type. The virtual machine’s graphics are accelerated for performance using the NVIDIA t4g graphics processor along with hardware-based encoding. The result is a seamless streaming of the user interface to the consumer’s browser.

Figure 1. vSkipGen cockpit container stack on AWS Graviton processor

Figure 2 below illustrates how Panasonic Automotive’s solution helps development teams to use the same host instance to run multiple isolated vSkipGen containers in parallel, providing a cost-optimized solution for large teams. Dynamic sizing of vCPUs and RAM helps teams to configure the vSkipGen environment so they can start experimenting immediately with the computational power of software-defined vehicles. In the same way, teams can virtually restrict resources to test the system’s boundaries.

Figure 2. vSkipGen’s host instance with support containerization on AWS Graviton processor

Benefits of vSkipGen in the software development lifecycle

DEVELOP FAST

• Achieve high-level environmental parity

vSkipGen achieves an operating-system level of environmental parity through instruction code compatibility and by verifying compatibility with the virtIO. Developers can agree on the interfaces that need to be made available in the next release and concurrently move forward, increasing productivity without needing to wait for a bench or to flash firmware on physical devices between stable and unstable versions.

• Enable frequent software iteration

The implementation of fully automated CI/CD/CT pipelines for automotive embedded software facilitates frequent iteration. These automated pipelines provide the consistent testing of every code change, thereby reducing the likelihood of human errors and enhancing software quality.

DEVELOP COMPETITIVELY

• Optimize Developer Costs for Virtual Environments

Developers can efficiently optimize resource allocation by utilizing smaller instances for tasks that do not necessitate a user interface, thereby reducing costs. This approach enables parallel development, allowing developers to address different aspects of the framework using the most suitable instance types, ultimately expediting development timelines. Panasonic has enabled up to 10 developers to use one G5G instance concurrently, drastically reducing the hourly cost for cloud-based development.

• Increase Sustainability through Resource Usage Optimization

Cloud-based development is more sustainable due to fewer physical parts that require production, shipping and maintenance. As software complexity increases, the likelihood of growth across teams is high, therefore translating to increased needs for prototype hardware for future vehicle development. The use of vSkipGen can help automakers optimize their logistics accelerating development velocity. This approach allows users to fit to their exact needs and scale as required.

DEVELOP CONFIDENTLY

• Support for Endurance Testing and Vehicle Simulations

Increasing test mileage through simulated vehicle rides offers valuable benefits at scale. This use case involves running simulations of vehicles traversing diverse routes, with the flexibility to generate simulations on the fly, use templates, or replay recorded scenarios. The advantages include achieving high-test coverage by running numerous instances for extended durations, effectively identifying edge cases and potential issues often missed in real world testing, such as memory leaks over longer durations. Simulated testing is more cost-effective it eliminates fuel and maintenance expenses associated with physical vehicles, as well as it frees up precious test driver times.

• Validation Test Parallelization and Automation at Scale

Scalability within this context facilitates the parallel testing of code changes, resulting in expedited feedback to developers and the swift identification and resolution of issues. Moreover, the automation minimizes downtime typically associated with manual testing, facilitating uninterrupted development and the seamless integration of new features. Additionally, the scalability of testing environments helps to run extensive regression tests, verifying that new code changes do not introduce unexpected issues and ultimately bolstering software reliability.

• Easy Issue Reproduction and Advanced Debugging

The storage of system configurations, development and production releases replication of system states streamlines debugging and issue resolution by providing rapid access to specific system conditions. It enhances testing and validation efforts by offering a diverse range of system states for comprehensive testing, including standard and edge cases, thereby ensuring that software updates perform reliably. This approach enables collaboration efficiency among geographically distributed teams by ensuring consistency in testing and troubleshooting processes through shared system states.

• Business continuity planning advantage

In addition to offering identical binary capability, the Virtual platform empowers teams to thoroughly assess hardware and resource requirements for product development. With the industry increasingly embracing ARM architecture for target platforms, OEMs gain the freedom to select desired SoCs without concerns about seamless software integration. This hardware-agnostic approach fosters a robust ecosystem and ensures a level playing field for all players, particularly OEMs, who can confidently choose different platforms for various model years, knowing that the software remains consistent and fully portable.

Conclusion

In this blog, we demonstrated how automakers can improve vehicle software quality and overcome hardware constraints using Panasonic Automotive’s solution vSkipGen, built on AWS. In vSkipGen, developers use virtual targets for software development, sidestepping the need for physical hardware while accessing a browser-based interface that streams the user interface through WebRTC. This approach is particularly advantageous for globally distributed teams and enhances cost-efficiency by removing a significant portion of hardware expenses. If you are interested in using vSkipGen today, please visit Panasonic vSkipGen on AWS Marketplace to learn more.