#Japan #AI Technology #Tokyo #TIER IV #Autoware

Introduction

Tier IV, a Tokyo-based deep tech company, is leading the charge toward democratizing autonomous driving technology. The company has unveiled an AI-driven software stack designed for Level 4 autonomous driving, aiming to maximize its operational design domain (ODD). This software stack is accessible through the open-source software repository, Autoware, and is adaptable to various systems-on-chip (SoC) and sensor configurations based on the automotive industry's needs. Additionally, Tier IV's Machine Learning Operations (MLOps) will empower automotive manufacturers to continually enhance their AI models using real driving data. To validate the effectiveness of this software, test drives are being conducted in Tokyo, Pittsburgh, and Munich in collaboration with local universities.

Software Stack Launch

Tier IV proposes the concept of "Level 4+" autonomous driving, aiming to extend the application of complete autonomous functionality to diverse driving environments incrementally. This new concept begins with the established capabilities of Level 4 in certain conditions and leverages data gathered from real-world operations to iteratively improve AI performance and expand use cases. At the core of this proposition is the newly released data-centric AI technology.

The software stack, building on the E2E architecture released in July 2025, expands Autoware's functionality. It is designed to be adaptable in diverse driving environments and to work flexibly across different hardware configurations without relying on any specific setup. Users can choose from two main software configurations:

1. Hybrid System: This combines recognition AI and path-generating AI. It utilizes diffusion models to probabilistically capture changes in the surrounding environment over time, integrating this with outcomes from another machine learning model to simulate human-like driving behaviors and trajectories.

2. End-to-End (E2E) System: This approach treats the environment and driving conditions in vector form, employing the concept of a world model. It aims to learn perception, judgement, and action in an integrated manner, achieving a seamless process from environment recognition to driving decisions and trajectory generation.

Collaboration with Open Source

The software stack is managed through GitHub in collaboration with The Autoware Foundation, which ensures fair development oversight. Through this framework, involving academia, industry, and community, Tier IV seeks to establish AI-based Level 4 autonomous systems as industry standards while advancing cutting-edge technologies.

Global Testing Across Regions

To evaluate the efficacy of this data-centric AI in various regions, Tier IV is initiated test drives that adhere to local regulations and safety standards across Japan, the United States, and Europe. Each test lasts approximately 60 minutes, with safety personnel accompanying the vehicle without intervening during operations deemed safe.

• - In Tokyo, in collaboration with the University of Tokyo, the team is validating the user experience while using Toyota's 'JPN TAXI' to navigate between urban hubs.
• - In Pittsburgh, tests with Carnegie Mellon University involve using Hyundai's 'IONIQ 5' for proof-of-concept robot taxis in city environments, including routes to and from Pittsburgh International Airport.
• - In Munich, tests with the Technical University of Munich will employ the Volkswagen 'T7 Multivan' to assess safety across various driving scenarios around the campus.

Enhancing ML through MLOps

Tier IV's MLOps facilitates quality checks for driving data, anonymizes individuals captured in footage, and offers a tagging system for effective data retrieval. It leverages active learning for data annotation and integrates virtual data generated under various conditions to evaluate autonomous driving functionality. This cutting-edge development is continuously supported through partnerships, including collaboration with Matsuo Laboratory.

In future implementations, Tier IV aims to continue enhancing AI performance by making use of extensive driving data and numerous functions offered by MLOps in collaboration with automotive manufacturers.

Comments from Leadership

Shintaro Kato, CEO of Tier IV

Kato emphasized the necessity for adaptive self-improving technology tailored to driving environments, framing the new data-centric AI model and cooperative machine learning infrastructure as a universal language that can serve as a starting point for the whole industry. He stressed that collaboration through "Autoware" will advance the development of autonomous driving technologies, ensuring it's not confined to specific organizations but instead opens pathways for societal benefit and progression.

Yang Zhang, Chairman of The Autoware Foundation

Zhang highlighted that "Autoware" acts as a foundation for the global collaboration of researchers, companies, and developers aiming to advance autonomous software technologies. The partnership with Tier IV will fortify international collaboration frameworks for continuous verification and refinement of E2E systems through real-world testing across Japan, the U.S., and Europe.

Prof. Yutaka Matsuo, University of Tokyo

Matsuo regarded the release of this software stack and MLOps as a significant leap towards implementing sophisticated AI models for industry purposes, aiming to bridge data accumulation from unique Japanese traffic conditions back to the development of "Autoware."

Raj Rajkumar, Carnegie Mellon University

Rajkumar acknowledged that "Autoware" is crucial in shaping Level 4+ autonomous driving concepts, with ongoing validations reflecting the technology's effectiveness under urban traffic conditions.

Prof. Johannes Betz, Technical University of Munich

Betz expressed excitement about this opportunity to evaluate Level 4 autonomous technology in European urban settings, expecting the framework based on "Autoware" will significantly enhance practical autonomous technology.

About Tier IV

Tier IV, established in December 2015, is dedicated to the vision of democratizing autonomous driving. As the lead developer of the open-source autonomous driving software 'Autoware', Tier IV provides a robust software platform aimed at meeting market demands through various autonomous driving services. The company strives to expand the possibilities of autonomous systems in partnership with stakeholders globally, promoting a safer and more sustainable society.

For press inquiries, contact: [email protected]