Autonomous Driving AI Training Patent Map

Examples of patents from Toyota, NVIDIA, and Bosch

2026/10/26

Summary
Overall trends
Collection, sorting, and labeling of real-world driving data
Patent case study: Toyota Motor Corporation
Training using simulation and synthetic data
Patent case study: NVIDIA Corporation
Vehicle-side learning, distributed learning, and update model distribution
Patent case study: Robert Bosch GmbH

Summary

This report was created by MarkLines based on patent information from the "Technical Information Distribution Service - AI Ninja" provided by Toyota Technical Development Corporation (TTDC), and covers current technology topics surrounding the automotive industry. TTDC operates an intellectual property (IP) business and a measurement and control business. The intellectual property business collects and analyzes information on global automobile development, provides consulting on research projects, and handles foreign patent applications and technical translations.

In recent years, as autonomous driving and driver-assistance technologies have become more sophisticated, development has progressed in AI used for recognizing the surrounding environment, predicting the behavior of other vehicles, and making driving decisions. To improve the performance of such AI systems, it is important to incorporate a wide variety of traffic situations into the training process. Conversely, during actual driving, it is difficult to gather sufficient data on low-frequency situations, such as those that could lead to accidents, adverse weather conditions, or others. For this reason, in addition to technologies that identify and prepare scenarios useful for training based on real-world driving data, technologies that supplement training scenarios using simulations and synthetic data are also being utilized. Furthermore, mechanisms are being considered for additional training on the vehicle side, and for reflecting the updated model based on that training back into the vehicle.

The number of patent applications related to the training of autonomous driving AI has increased from 2020 to 2025. By technology category, the highest number of patent applications relates to “Collection, sorting, and labeling of real-world driving data,” followed by “Training using simulation and synthetic data” and "Vehicle-side learning, distributed learning, and update model distribution." In addition, from 2024 to 2025, while the number of cases involving “Collection, sorting, and labeling of real-world driving data” has decreased, the number of cases involving “Training using simulation and synthetic data” as well as “Vehicle-side learning, distributed learning, and updated model distribution” has increased.

Based on this background, this report examines patent applications related to the training of autonomous driving AI technology in Japan, the United States, China, Europe, South Korea, and international applications (WO), organizing them from three perspectives: "Collection, sorting, and labeling of real-world driving data," "Training using simulation and synthetic data," and "Vehicle-side learning, distributed learning, and updated model distribution architectures" represent the cutting edge of the software-defined vehicle (SDV) industry. The first perspective deals with technologies for collecting and sorting data obtained from actual vehicles and roads into a format suitable for training, and for labeling that data with the correct information. The second perspective focuses on technologies that use driving data created and processed using virtual environments and generative technologies for AI training. The third perspective addresses technologies related to training on individual vehicles or across multiple vehicles, as well as the distribution and application of trained and updated model architectures to vehicles. We will analyze trends in the number of applications and the activities of major applicants to gain an overall understanding of patent applications in this technology field.

* The number of patent applications for 2025 is a provisional figure, as many have not yet been made public.

Toyota Technical Development Corporation

Technical Information Distribution Service - AI Ninja URL: https://thinktank.toyota-td.jp/pub/list
Contains much content related to automobiles. Graph the content you're interested in with just one click. In addition, detailed display and downloading are also supported.

This report is for paid members only.

If you register as a free member, you can read the rest of this article for a limited time.
In addition, you can also enjoy the following content for free:

Market & Tech Reports

Global automotive production/sales

Launch schedule forecasts