TIER IV and the Autoware Foundation announced Autoware 2.0 this week, a major upgrade that adds a selectable AI planning option while keeping the established rule‑based stack as the default. TIER IV said it will lead industry initiatives to integrate Autoware 2.0 with partners including Arm, AWS and Red Hat as part of a push to accelerate commercial adoption. (Announcement date: Nov 2, 2026.)
What Autoware 2.0 changes
Technically, version 2.0 introduces an architecture the project calls a "generator‑selector" framework: multiple trajectory generators (rule‑based or learned) produce candidate paths and a selector performs safety checks and chooses the final trajectory. In this first 2.0.0 release developers can launch Autoware with the existing rule‑based planner (the default) or with the new Diffusion Planner by setting the runtime argument to "diffusion_planner"; the release notes also show model and dataset hosting moved to Hugging Face with pinned version tags. The Autoware documentation explains that the release ships with the Diffusion Planner using model tag v5.0 and that selection is decided at launch.
TIER IV framed the release as part of a broader ecosystem play. In its Nov. 2 statement the company said Autoware 2.0 "expands support for data‑centric, AI‑based approaches while maintaining compatibility with conventional rule‑based approaches," and announced collaboration plans with Arm, AWS and Red Hat to align compute platforms, cloud workflows and the AutoSD upstream operating system project. Shinpei Kato, founder and CEO of TIER IV, said: "Autoware 2.0 represents an important step toward the democratization of autonomous driving in the era of AI‑defined vehicles." Muhammad Zain Khawaja, president of the Autoware Foundation, welcomed the wider community effort behind the release.
