Toyota Research Institute - Advanced Development, Inc.

TRI-AD was founded in March 2018 via a joint investment between Toyota Motor Corporation, Denso, and Aisin Seiki. Through the advanced development of software for automated driving, the company has combined research and mass production to pioneer the field of automated driving.

Thanks to the development of control systems and AI technology, the processes that govern “decisions” and “operations” have evolved rapidly. On the other hand, when considering the daily changes to roads, it is no easy task to maintain mapping information and ensure that it remains up to date. According to Mandali Khalesi, VP of the Automated Driving division, “these difficulties have meant that maps are presently limited to specific nations and regions, or to high-speed roadways where the collection of up-to-date information is comparatively easy.”

However, carrying out continual updates to maps of public highways with high precision has proved a major burden when carried out by a single corporate group. In response, TRI-AD announced the construction of an “Automated Mapping Platform (AMP)” in January 2019, one in which any company could participate in the creation and use of high definition maps.

AMP uses information gathered from satellite images and sensor data from automobile makers such as Toyota Motor Corporation, taxi companies, and delivery companies. Such information allows for the efficient creation and update of high-precision maps. AMP enables developers at automobile makers and suppliers to use the maps, which has accelerated development of Toyota Motor Corporation’s automated driving applications.

Khalesi continued by stating “At present, a number of vendors who possess satellite imagery and onboard camera technology, from automotive companies within the Toyota Motor Group to those outside, have expressed their intention to participate in proof of concepts (or “PoCs”) and we expect more partners to join us in the future.”

After the announcement of its new AMP architecture, TRI-AD launched a prototype to carry out PoCs with partner companies. In order to prioritize speedy development, AWS was adopted early with the intention of exploring the use of cloud services. AWS is already being used for a number of projects at TRI-AD, including machine learning and deep learning. It has become one of the standard elements of the company’s IT infrastructure. AWS is also used frequently by Toyota Research Institute (TRI), an AI technology research organization in Silicon Valley, and both companies have experienced the merits of its development efficiency. Not only that, but use of the management services offered by AWS was said to be another deciding factor in its adoption. Ryo Igarashi, the engineer in charge of AMP’s cloud operations development, stated that “given the limited staffing resources that confront developers, the rich management resources AWS offers have been essential for achieving short-term development. In addition, we expect the number of partner companies participating in PoCs to increase and even if new features are developed, the scalability that enables easy expansion of resources and services has been highly rated. ”

According to TRI-AD’s design, AMP consists of three elements: collection and accumulation of data, highly precise map development, and the broadcasting of these highly precise maps. The core of data collection and accumulation consists of two segments, the real-time upload of data from running automobile probes and the batch upload of compiled data from taxi companies and other fleet partners. In terms of map development, developers from companies participating in AMP can access the Developer Portal and create map information tailored to their application by developing and deploying unique algorithms in the cloud. The resulting data will be registered and stored in a specialized Map Marketplace and broadcasted as an API, as necessary, to their autonomous fleet.

“Our AMP prototype is based on the policy of first making the product and then trying it out, meaning agile development where improvements are made as it is used. By prioritizing an environment where PoCs with partners can proceed easily, we focused on scalability that would be necessary for its future usage," says Igarashi.

Development of the prototype began in April 2019, and it was released two months later in June. By coordinating with the architects of AWS solutions and professional service consultants, TRI-AD’S developers were able to achieve a shorter development period.

The prototype was created via active utilization of serverless architecture. Use of client certificates issued by AWS IoT formed the basis for real-time upload of data collection and accumulation tools. Tokens were received from AWS STS and certified by the Amazon API Gateway, ensuring the secure upload of data. Batch uploads were based on user certification by Amazon Cognito, while the running vehicle data and metadata were collected via Amazon S3 and AWS AppSync, respectively, before being accumulated on Amazon DynamoDB.

"As a next step, we will begin work on the architecture of the future usage environment with the goal of implementing services," Khalesi said. "In pursuit of future global expansion within America, Europe, and Asia, we are planning to build in overseas compatibility and strengthen coordination with our international partners. AMP will enable rapid innovation in the field of automated driving and, together with our partners, will contribute to solving social issues.”