Toyota Racing Brings Data Together in the AWS Cloud to Win Races

NASCAR has seen something of a technological revolution in recent years, with Toyota Racing Development (TRD) leading the way. Over the past 5 years, TRD has used insights gleaned from historical competition data and current vehicle data to help the company achieve NASCAR Cup Series manufacturers’ championships in 2016, 2017, and 2019. But with other racing teams in hot pursuit, TRD sought to press its technological advantage by building a data repository that would bring the company’s siloed applications into one centralized location, enabling the TRD team to find the important information it needed to make snap decisions during races. TRD, which had relied on Amazon Web Services (AWS) for its initial technological push in 2015, again called on the cloud provider to bring data together in its core data platform (CDP). Among the more than 40 AWS services TRD uses are Amazon Athena, an interactive query service that makes it easy to analyze data in Amazon Simple Storage Service (Amazon S3) using standard SQL, and AWS Lambda, which enables TRD to run code without provisioning or managing servers.

A subsidiary of Toyota Motor North America, TRD competes across multiple platforms, including the NASCAR Cup Series. “Our primary goal as a company is to win,” says Jonny Elliott, TRD’s senior engineering manager of technology. And that’s exactly what TRD has done, winning three of the last five NASCAR Cup Series manufacturers’ championships and three of the last five drivers’ championships. TRD’s journey to its first NASCAR Cup Series championship began in 2015, when the team allotted more resources to software engineering. The first application TRD built was a timing and scoring application to track the leaderboard and lap times during races. TRD would go on to build more than 20 applications that collected practice, qualifying, and race data. “Over the years we had accumulated an awful lot of data,” says Elliott, “but it was disparate, and it wasn’t really providing value. So that’s why we moved toward the CDP.”

Creating a centralized repository would help TRD connect siloed data. Instead of spending crucial time looking in different buckets for key information—including engine data, race images, and brake data—the TRD team sought to bring it all together to make it more accessible. Having grown in the cloud alongside AWS over the years, TRD decided to bring in experts from AWS Professional Services to build the CDP starting in January 2020. “We worked on AWS pretty much from the start,” says Elliott. “Everything we’ve built, we’ve built from scratch, and we’ve built it on AWS.”

For phase one of the project, TRD and the AWS Professional Services team examined TRD’s existing data infrastructure and got to work moving racing data into the CDP. “It took 3 or 4 months to build the infrastructure for the CDP and be ready to go,” says Elliott. “And then we moved on to the next application. So far we’ve moved through four or five applications, and that got us to the end of phase one.” During races, TRD expects to use the centralized data to make crucial in-race decisions quickly.

“Everything is about speed and trying to get the data in people’s hands as quickly as possible,” says Elliott. “Let’s say a caution comes out on a certain lap because somebody crashed into your car. Your driver tells the crew chief, ‘I got hit,’ but he’s a mile and a half away, and the crew chief can’t see anything. With everything linked up in the CDP, the crew chief doesn’t have to scroll through 5,000 pictures taken from a trackside photographer; we can put the right images in his hands, and he can decide whether to bring the car into the pits. The difference between 3 seconds and 30 seconds could be the difference between missing a lap or getting the driver in.” TRD is ready for this scenario in part because of Amazon Rekognition, which makes it easy to add image and video analysis to applications using proven, highly scalable deep learning technology that requires no machine learning expertise to use. Using Amazon Rekognition, the team is able to tag images of the car and the time, facilitating instant access to the correct photo within the correct application.

To query data in the CDP, the team uses Amazon Athena. Amazon DynamoDB, a fully managed secure database, serves as the database backbone for a variety of the company’s applications. Additionally, TRD has been updating applications by integrating AWS Lambda. “I’d say 80 percent of our applications are using AWS Lambda to some degree,” says Elliott. “Things like AWS Lambda have been a driver for changing how we do things.”

Another critical application involves simulations, which TRD runs to optimize its cars’ on-track performance. Previously TRD ran simulations on clustered physical servers that took hours to complete—often overnight. Once TRD turned to Amazon Elastic Compute Cloud (Amazon EC2), a web service that provides secure, resizable compute capacity in the cloud, it cut that time down to 15 minutes. Bringing these fast simulations into the CDP enabled TRD to run them with data from practice laps on race day and make immediate use of the results. The TRD team also can spin up Amazon EC2 c5.9xlarge and c5.18xlarge instances when it needs to run simulations and then spin them down again when it doesn’t, reducing the costs associated with unused compute capacity and thereby saving about $1 million annually. “Being able to run hundreds of thousands of simulations very quickly in the cloud was a massive gain for our race teams,” says Elliott.

Bringing race data together is just the beginning of TRD’s ambitions for its CDP. Phase two involves using data science and machine learning to recognize patterns in the data and make race-time predictions. The TRD team wants to be able to predict things like the occurrence of a yellow flag during a race or to play out the relative probabilities of winning based on the behavior of the other cars on the track. Part of this solution involves Amazon SageMaker, a fully managed service that provides every developer and data scientist with the ability to build, train, and deploy machine learning models quickly. TRD saw Amazon SageMaker in action in the AWS DeepRacer League, the world’s first global autonomous racing league. “It was really good for my team of software developers to get some exposure to Amazon SageMaker and machine learning in a fun, competitive way,” says Elliott.

TRD was able to build a series of critical racing applications on AWS and bring all relevant data together in order to make snap decisions during races. TRD’s savvy adoption of cloud technology is one reason why TRD keeps on winning despite having fewer race teams than its competitors. “It’s quality over quantity,” says Elliott. “We put an awful lot of effort and emphasis into the small amount of people that we have to try to make them elite. We’re confident that the software that we’ve produced over the past 5 years and the tools that we’ve provided our race teams are some of the best in motorsports.”