#Japan #AI Solutions #Tokyo #Autonomous Driving #TIER IV

TIER IV Showcases Integrated AI and Data Solutions for Software-Defined Vehicles at Automotive World 2026

TIER IV, a trailblazer in the realm of open-source software for autonomous driving, is gearing up to make waves at the upcoming Automotive World 2026, scheduled for September 9-11 at Makuhari Messe, Tokyo. The company plans to unveil a new integrated solution targeting software-defined vehicles (SDVs), providing a synthesis of artificial intelligence, data, and computing capabilities.

What to Expect at the Event

During its showcase, TIER IV will introduce the autolabeling feature of its Co-MLOps platform, which facilitates efficient data sharing and processing. This will be accompanied by a demonstration of the reference end-to-end (E2E) AI model—an innovative platform designed explicitly for autonomous driving. The highlight of this demonstration will be running the E2E AI model on an automotive computing framework, thereby illustrating an all-encompassing integrated SDV solution that spans from data acquisition and processing to AI model development and deployment on the automotive computing system.

Importance of AI in Autonomous Driving

The drive toward integrating AI within autonomous driving technology is continuously gaining momentum. The ability of AI to accurately interpret and interact with its environment hinges upon access to diverse and high-quality datasets. These datasets must come with accurate labeling, necessitating the establishment of efficient frameworks to develop effective autonomous driving AI models. To facilitate the mass adoption of these AI solutions within passenger vehicles, their efficient integration within automotive computing platforms is critical.

In response to these challenges, TIER IV launched the Co-MLOps platform in January 2024. This initiative is focused on collaborating with partners to gather and disseminate driving data. TIER IV aims to streamline the development of an integrated SDV solution that brings together AI, data, and computing by employing an autolabeling function that simplifies the complex task of data labeling for driving data. This will enhance the efficiency of creating the E2E AI model and establish a solid infrastructure for its deployment.

Key Aspects of the Autolabeling Function

To leverage driving data as training material for autonomous driving AI models, a substantial amount of accurate and well-labeled data is required. The autolabeling function is designed to automatically annotate a wide variety of objects and environment elements within driving data—such as vehicles, pedestrians, and infra-related structures—ensuring they can be promptly recognized by AI systems.

This innovative feature can produce millions of labels almost instantaneously while maintaining high-quality consistency, which dramatically increases data processing efficiency, even as more sensors are introduced or data volumes rise. By alleviating the burden of manual labeling, TIER IV anticipates a reduction in development costs and a subsequent increase in the speed of autonomous driving AI model production.

To further bolster the accuracy of these AI models, TIER IV complements difficult-to-gather driving data—like rare scenarios that present collision risks and atmospheric conditions—by incorporating synthetic data generated using NVIDIA Cosmos. The autolabeling feature also applies to this supplementary data.

Features of the Reference E2E AI Model

The reference E2E AI model does not depend on high-definition mapping; instead, it utilizes images from automotive cameras to conduct various functions—from ascertaining a bird's-eye view of the surrounding environment to generating vehicle trajectories. All these functions can be executed via a single neural network. The model is capable of identifying 3D roads and objects, generating occupancy maps to determine potential obstacles, and predicting vehicle movements. Additionally, its design aligns well with automotive computing, facilitating its deployment across real-world system-on-chip implementations.

The development of this model employs an agent-driven AI process where human inputs dictate the construction parameters, while an AI agent spearheads the development procedure—leveraging automatically labeled data, optimizing implementations, and managing models. This autonomous overseer manages the cycles of training, testing, and improvement, ultimately enhancing development efficiency and enabling rapid prototyping at lower costs.

TIER IV continues refining the reference E2E AI model, with plans to offer it to expectant partners in the Co-MLOps initiative.

Live Demonstration at Automotive World 2026

At the Automotive World 2026 exhibition, TIER IV will execute a live demonstration showcasing the capabilities of its autolabeling function. Attendees will observe how this feature identifies various objects and elements within a driving context. The demonstration will highlight the reference E2E AI model operating on an NVIDIA Jetson Orin, illustrating the model's ability to navigate its environment and generate vehicle trajectories solely from input images collected by automotive cameras.

By harmonizing extensive data collection and processing with continuous enhancement of the E2E autonomous driving AI model, TIER IV is making significant strides toward an SDV solution intended to assist automakers and suppliers in innovating autonomous driving capabilities for the mass market.

About TIER IV

TIER IV is a leader in deep tech innovations and is responsible for pioneering Autoware, an open-source software suite for autonomous driving. Offering a broad range of platforms and services centered around Autoware, TIER IV supports everything from software development and vehicle acquisition to operational logistics. Through collaboration with partners globally, TIER IV is actively shaping the future of intelligent mobility through open-source solutions, striving to create transportation options that are safer, sustainable, and sufficiently accessible for everyone.