Turing Begins Joint Research on In-Vehicle E2E Systems and Physical Foundation Models with Automotive Firms

Turing Inc. (Ota-ku, Tokyo; CEO: Kazunari Yamamoto, hereafter "Turing") announces that it has begun joint research with SUBARU Corporation (Shibuya-ku, Tokyo; President and CEO: Atsushi Osaki, hereafter "SUBARU") and DENSO Corporation (Kariya, Aichi; President and CEO: Shinnosuke Hayashi, hereafter "DENSO") on in-vehicle End-to-End (E2E) autonomous driving systems and physical foundation models.

In this joint research, (1) Turing and SUBARU will focus on research toward establishing in-vehicle E2E autonomous driving system technology, while (2) Turing and DENSO will focus on the development of physical foundation models. By consolidating the expertise of each company, Turing will construct a system to promote the development, from AI model research to implementation in real vehicles, in an integrated manner.

Background For the social implementation of full autonomous driving, it is essential to have a development framework that iteratively conducts the development of physical foundation models that perform high-level recognition and judgment based on large-scale driving data, and initiatives to mount and verify them in real vehicles. Turing has been advancing the in-house development of E2E autonomous driving and physical foundation models. In this joint research, Turing aims to collaborate with both companies, which have supported the automotive industry for many years, to build a framework that promotes the entire process from research and development to implementation in real vehicles and public road testing.

Key Initiatives This joint research will proceed with the following two initiatives in parallel:

1. In-Vehicle E2E Autonomous Driving System Working with SUBARU, the project will focus on the integration of lightweight E2E autonomous driving models with vehicle control systems, with an eye toward future public road testing. The team will advance the development of lightweight E2E models using multiple cameras and their adaptation for vehicles, constructing and verifying an autonomous driving system that includes vehicle control.

2. Physical Foundation Models Working with DENSO, the project will focus on the development of physical foundation models tailored for autonomous driving and their adaptation to real vehicles. Based on visual-language models and visual state-space models, the team will develop technology to acquire linguistic and spatio-temporal understanding from camera sensors alone.

Future Outlook Turing aims to accelerate the R&D of E2E autonomous driving systems and physical foundation models through this joint research, and to build a development framework that cycles through deployment to real vehicles, verification in driving environments, and model improvements based on the driving data obtained.

Through these efforts, Turing aims to establish a foundation for the social implementation of full autonomous driving and connect this to the establishment of physical AI technology that can be applied to a wide range of fields, including multimodal foundation models for AI robots, beyond just autonomous driving.

Project Positioning This project is implemented as part of the "Post-5G Information and Communication System Infrastructure Enhancement R&D Project / Development of Competitive Generative AI Foundation Models (GENIAC)" promoted by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO).