Open AD Kit: The First SOAFEE Blueprint Powering Open Collaboration in Autonomous Driving
The Automotive Industry’s Software Turning Point
The automotive industry has entered an era of radical reinvention. Multiple technological and societal shifts are converging — from the rise of electrified powertrains and unified vehicle compute architectures to the rapid progress of intelligent driving systems and personalized, ever-evolving in-vehicle experiences. Vehicles are no longer static machines defined at production; they are becoming dynamic digital platforms that continue to evolve throughout their lifetime.
This marks the emergence of the software-defined vehicle, where the true value of mobility is created and sustained through software. Other industries have already proven the impact of this shift. Software-defined data centers, software-defined networking (SDN) in telecommunications, and software-defined systems in aerospace and manufacturing have demonstrated how shared, flexible software layers can transform entire value chains. The automotive sector has now reached the same threshold — ready to harness software-driven innovation to achieve new levels of agility, interoperability, and scalability.
For both established and emerging automakers, this transformation is redefining how vehicles are conceived and delivered. New players are moving fast with software-first mindsets and clean-sheet architectures, while traditional OEMs modernizing deeply rooted systems and development processes. What unites both is the recognition that software is now the primary differentiator — yet much of the stack beneath it, particularly between the hardware and application layers, is non-differentiating. The industry’s real advantage lies in collaborating on those common layers, building on open and standardized foundations that reduce complexity and duplication of effort. Collaboration at this level enables every player to focus their innovation where it truly matters: at the application layer, where advanced capabilities such as autonomous driving, intelligent safety systems, and personalized mobility experiences are created.
Collaboration reaches its full potential when it happens in the open. Open source transforms collaboration from coordination into co-creation, enabling companies, researchers, and developers to build, test, and evolve technology together at a global scale. This mindset has already reshaped autonomy through projects like Autoware, where a shared, community-driven foundation accelerates innovation across the ecosystem.
We are showcasing in this blog how the Open AD Kit (The First SOAFEE Blueprint) is bringing that collaborative innovation to the SDV community – in a very tangible way.
Autoware: Open Collaboration at Full Scale
Nowhere is the spirit of open collaboration more evident than in Autoware — the world’s first open-source software stack for autonomous driving. For over a decade, the Autoware community has pushed the boundaries of open development, tackling one of the most complex workloads in the automotive domain: full vehicle autonomy. What makes it even more remarkable is that this innovation is happening entirely in the open, under the Apache 2.0 license, enabling both academic exploration and commercial adoption.
August 2025 marked Autoware’s tenth anniversary — ten years of global collaboration that have made it one of the most trusted and widely deployed autonomy platforms in the world. Today, OEMs, Tier 1 suppliers, research institutes, and universities all rely on Autoware — using it to build, prototype, benchmark, and validate autonomous driving systems, often alongside or in comparison with proprietary ADAS and autonomy stacks.
Autoware isn’t just a package or middleware — it’s a full-stack autonomy solution, from raw sensor data ingestion to vehicle actuation. It powers a wide range of real-world use cases, including autonomous shuttles and buses, robotaxis, low-speed and special-purpose vehicles, racing platforms, and even efforts aimed at bringing automated driving into privately owned vehicles. Its maturity and credibility are reflected in its thriving ecosystem, with a record number of forks, deployments, and contributors — representing more than 10,000 developers who have built, tested, and advanced autonomy through open collaboration.
It was within this same spirit of openness and collective innovation that the Open AD Kit was born — a foundational effort to extend Autoware’s principles into the broader software-defined vehicle ecosystem.
The Open AD Kit Blueprint: A Collaborative Ecosystem in Motion
As Autoware matured, its modular design naturally evolved toward a microservices architecture, enabling components to be developed and tested independently. This provided the perfect foundation to explore hardware abstraction via containerization — the seed of the Open AD Kit.
Early experiments were strikingly successful. Once containerized, Autoware workloads could run not only on embedded edge hardware but also on cloud instances with system parity. For the first time, an autonomy stack traditionally built and deployed solely on in-vehicle compute could operate seamlessly in the cloud. This created a true A-ha! moment for key stakeholders — including Arm, AWS, and others — proving that autonomous-driving software could be developed and tested the cloud-native way, with the potential for continuous, perpetual updates.
The Open AD Kit Blueprint is co-developed by three consortia — the Autoware Foundation, SOAFEE, and the eSync Alliance — to provide a reference architecture that shows how cloud-native autonomy can be deployed, updated, and maintained across heterogeneous compute.
Within this blueprint, Autoware serves as the primary workload — a fully containerized autonomy stack built and maintained by Autoware Foundation member organizations and individual contributors. The eSync Alliance, through member implementations, powers the OTA pipeline between Arm-based cloud instances and Arm-based vehicle ECUs, deploying server/client/agent components in implementation demos and proof-of-value projects.
The blueprint brings together a broad ecosystem — from chipmakers and software-infrastructure providers to system integrators, vehicle developers, and mobility operators — who validate and extend the architecture through demos and direct projects with the Autoware Foundation.
What’s Inside the Blueprint — and Who Built on It
The Open AD Kit Blueprint began by containerizing planning and control (plus visualization), then increased granularity to include perception, sensing, and localization. This modularization enabled teams to build, test, and update individual subsystems independently — accelerating iteration across the autonomy pipeline.
The most frequently showcased example is the planning demo: an autonomous vehicle approaches a stationary obstacle, and the planning container is updated over the air to modify the takeover behavior. It is a concise illustration of software updateability without redeploying the entire stack. This demo has been widely adopted and presented by Arm (explore Arm Learning Paths), Amazon Web Services (AWS) (explore AWS’ Open AD Kit workshop), Red Hat (explore Red Hat’s SOAFEE Blueprint), Corellium (explore Open AD Kit running on Corellium), Synopsys, and others.
Open AD Kit also seeded multiple SOAFEE blueprints: DENSO built an Automated Valet Parking (AVP) blueprint using Autoware as the autonomy workload.
Red Hat created a SOAFEE blueprint integrating its upstream AutoSD Linux project, showing a development path toward the Red Hat In-Vehicle Operating System (RHIVOS) — a commercially supported, functional-safety-certified platform.
Each of these efforts reinforces the Open AD Kit’s purpose: a shared, open foundation the industry can build on to innovate, iterate, and deploy software-defined autonomy at scale.
What’s Next for Open AD Kit: From Blueprint to Living Platform
The Open AD Kit has always been more than a static reference — it’s a continuously evolving platform for the software-defined-vehicle era. As the ecosystem matures, upcoming milestones will focus on extending the blueprint into new dimensions of safety, AI integration, and deployment flexibility.
One major focus is safety-critical autonomy. The next phase of the Open AD Kit will showcase a Minimum Risk Maneuver (MRM) implementation that integrates a constrained safety island within a heterogeneous compute environment — combining
high-performance compute for perception and planning with safety-certified, constrained compute for fallback logic. This demonstration will highlight how modern SDV architectures can orchestrate mixed-critical workloads in a secure and deterministic way.
In parallel, the Autoware Foundation engineering teams are working to incorporate foundation AI models for end-to-end autonomous driving directly into containerized workloads. These models will simplify deployment and orchestration by allowing developers to mix and match perception, planning, and control pipelines with flexible sensor configurations, creating a faster path to experimentation and domain-specific optimization.
The team is also implementing a series of autonomy reference designs in the form of modular Open AD Kit containers. These designs will enable configurable autonomy functions — allowing developers to tailor the software stack to different vehicle types, compute budgets, and safety requirements.
Some of these developments will be featured in our upcoming CES demonstration, offering an early glimpse into the next evolution of the Open AD Kit. Following CES, the community can expect the release of Open AD Kit v2, which will introduce new demos and workloads for developers to explore, experiment with, and build upon.
Looking ahead, the Open AD Kit roadmap will continue to evolve through a series of follow-up blueprints. These will demonstrate key advancements such as:
- Running Open AD Kit within safety-certified frameworks, showcasing mixed-criticality orchestration and safety-island implementations.
- Enabling bi-directional data pipelines through integrated MLOps workflows, connecting development, deployment, and continuous learning loops within the Open AD Kit framework.
- Each of these milestones reinforces the project’s original mission: to make autonomy development faster, safer, and more accessible through open collaboration.
