TIER IV 联合 Autoware 基金会发布 Autoware 2.0，支持端到端与混合驾驶栈

TIER IV announced on November 2, 2026 that it is spearheading Autoware 2.0 initiatives in collaboration with the Autoware Foundation and the global open-source community, the same day the foundation released version 2.0 of the open-source autonomous driving software with new support for end-to-end AI and hybrid driving stacks alongside established rule-based modules.

TIER IV described Autoware 2.0 as a significant evolution of open-source autonomous driving software, expanding support for data-centric, AI-based approaches while maintaining compatibility with conventional rule-based approaches. The company said key developments include a scalable and adaptive architecture supporting various configurations of AI-based and rule-based systems, a flexible indexing scheme for modular software integration, and a ready-made Reference Design framework to facilitate commercial adoption across different vehicles, applications, and operating environments.

“Autoware 2.0 represents an important step toward the democratization of autonomous driving in the era of AI-defined vehicles,” said Shinpei Kato, founder and CEO of TIER IV. “Open-source software alone is not enough to transform the automotive industry. We need an ecosystem that connects AI, computing platforms, cloud infrastructure, and automotive software into practical solutions.”

The initiatives build on TIER IV’s existing collaborations with Arm, AWS, and Red Hat. With Arm, TIER IV is advancing the development of computing platforms for Autoware 2.0 as a representative use case of Arm Total Design for Physical AI. With AWS, the company is advancing cloud-native development and operations environments for Autoware 2.0, building on its existing use of AWS cloud technologies. With Red Hat, it is advancing system integration of Autoware 2.0 with AutoSD, the open-source upstream project for Red Hat In-Vehicle Operating System.

TIER IV said it aims through these complementary initiatives to strengthen the technology ecosystem supporting Autoware 2.0 and to expand opportunities for its adoption and commercialization across the automotive industry. Muhammad Zain Khawaja, president of the Autoware Foundation, said TIER IV has played a leading role in advancing Autoware and that its continued commitment to open-source development and industry collaboration is essential to the growth of the ecosystem.

Inside the Autoware 2.0 Release

Autoware has traditionally been structured as a modular pipeline for sensing, perception, localization, planning, and control, with packages organized into two layers: Autoware Core, a curated set of stable, well-tested libraries, APIs, and interfaces maintained by the foundation, and Autoware Universe, which extends Core with additional features, experimental algorithms, and community contributions. Autoware 2.0 builds on this architecture rather than replacing it, bringing rule-based and AI-based technologies into a shared architecture.

A new generator–selector architecture allows multiple trajectory generators to run within the same system, with a selector choosing which output to use, so developers can compose rule-based modules, AI modules, or a hybrid combination into a single architecture for their use case. The release includes documentation of the new architecture and a simulator demonstration with rule-based modules and various end-to-end modules.

The release also changes how AI models and datasets reach a workspace, replacing a setup in which every model download was defined in one central Ansible file. The Autoware Foundation’s organization on Hugging Face is now the default home for foundation models and datasets, with version control, while manifests can also point to the foundation’s ML server, AWS S3, or third-party servers so existing artifacts keep working during the migration.

A new Autoware Index acts as a registry of links to Autoware-compatible repositories and Reference Designs, where community members can register their packages and users can discover them in one place. A command-line tool, the Autoware Index CLI, lets developers pick packages or a whole Reference Design together with their dependencies and writes them to an autoware-index.repos file in the workspace. Five Universe packages moved to their own repositories with this release, and tags were reorganized under labels such as “ai”, “reference-design”, and “e2e”.

The Autoware Reference Design framework combines a Deployment Profile, describing the vehicle and hardware configuration, operating conditions, and use case, with a Reference Software Stack, a published baseline that resolves to a concrete set of modules, model versions, parameters, interfaces, and platform requirements. Candidate stacks include rule-based planning, an end-to-end stack, and a hybrid stack. Reference Designs are grouped into application families such as low-speed shuttles, highway assistance, and urban robotaxis. Each design states its status, distinguishing intended capability from demonstrated results, and operating domains are described using ASAM OpenODD terminology.

“Users often ask us how to go from Autoware’s software to a complete autonomous vehicle,” said Ryohsuke Mitsudome, TSC chair of the Autoware Foundation. “Autoware 2.0 addresses this by introducing Autoware Reference Design with a documented blueprint for the entire system.”

Autoware 2.0 is available now on GitHub, with documentation at docs.autoware.org, package and Reference Design discovery through the Autoware Index, and foundation models on Hugging Face.

ASSC US 2026 Presentation

TIER IV will join Arm, AWS, and Red Hat at Automotive Software Strategies US 2026, scheduled for November 3–4, 2026, in San Jose, California, where representatives of the four companies are expected to present their respective approaches to advancing automotive software technologies and open-source ecosystems. Christian John, president of TIER IV North America, will present the company’s vision for accelerating the commercial adoption of Autoware 2.0 through open-source collaboration and global technology partnerships.