As one of many quickest sports activities on the planet, virtually all the pieces is a race in Method 1® (F1), even the pit stops. F1 drivers must cease to alter tires or make repairs to wreck sustained throughout a race. Every valuable tenth of a second the automotive is within the pit is misplaced time within the race, which might imply the distinction between making the rostrum or lacking out on championship factors. Pit crews are educated to function at optimum effectivity, though measuring their efficiency has been difficult, till now. On this publish, we share how Amazon Net Providers (AWS) helps Scuderia Ferrari HP develop extra correct pit cease evaluation strategies utilizing machine studying (ML).
Challenges with pit cease efficiency evaluation
Traditionally, analyzing pit cease efficiency has required observe operations engineers to painstakingly evaluation hours of footage from cameras positioned on the entrance and the rear of the pit, then correlate the video to the automotive’s telemetry information. For a typical race weekend, engineers obtain a median of twenty-two movies for 11 pit stops (per driver), amounting to round 600 movies per season. Together with being time-consuming, reviewing footage manually is liable to inaccuracies. Since implementing the answer with AWS, observe operations engineers can synchronize the information as much as 80% quicker than guide strategies.
Modernizing by way of partnership with AWS
The partnership with AWS helps Scuderia Ferrari HP modernize the difficult means of pit cease evaluation, through the use of the cloud and ML.
“Beforehand, we needed to manually analyze a number of video recordings and telemetry information individually, making it troublesome to establish inefficiencies and rising the chance of lacking vital particulars. With this new method, we are able to now automate and centralize the evaluation, gaining a clearer and extra fast understanding of each pit cease, serving to us detect errors quicker and refine our processes.”
– Marco Gaudino, Digital Transformation Racing Software Architect
The answer makes use of object detection deployed in Amazon SageMaker AI to synchronize video seize with telemetry information from pit crew tools, and the serverless event-driven structure optimizes using compute infrastructure. As a result of Method 1 groups should adjust to the strict price range and compute useful resource caps imposed by the FIA, on-demand AWS providers assist Scuderia Ferrari HP keep away from costly infrastructure overhead.
Driving innovation collectively
AWS has been a Scuderia Ferrari HP Group Accomplice in addition to the Scuderia Ferrari HP Official Cloud, Machine Studying Cloud, and Synthetic Intelligence Cloud Supplier since 2021, partnering to energy innovation on and off the observe. In the case of efficiency racing, AWS and Scuderia Ferrari HP often work collectively to establish areas for enchancment and construct new options. For instance, these collaborations have helped scale back car weight utilizing ML by implementing a digital floor velocity sensor, streamlined the energy unit meeting course of, and accelerated the prototyping of latest industrial car designs.
After beginning improvement in late 2023, the pit cease answer was first examined in March 2024 on the Australian Grand Prix. It shortly moved into manufacturing on the 2024 Japanese Grand Prix, held April 7, and now offers the Scuderia Ferrari HP workforce with a aggressive edge.
Taking the answer a step additional, Scuderia Ferrari HP is already engaged on a prototype to detect anomalies throughout pit stops robotically, equivalent to difficulties in lifting the automotive when the trolley fails to carry, or points in the course of the set up and removing of tires by the pit crew. It’s additionally deploying a brand new, extra performant digicam setup for the 2025 season, with 4 cameras taking pictures 120 frames per second as an alternative of the earlier two cameras taking pictures 25 frames per second.
Growing the ML-powered pit cease evaluation answer
The brand new ML-powered pit cease evaluation answer robotically correlates video development with the related telemetry information. It makes use of object detection to establish inexperienced lights, then exactly synchronizes the video and telemetry information, so engineers can evaluation the synchronized video by way of a customized visualization device. This automated methodology is extra environment friendly and extra correct than the earlier guide method. The next picture reveals the thing detection of the inexperienced gentle throughout a pit cease.
“By systematically reviewing each pit cease, we are able to establish patterns, detect even the smallest inefficiencies, and refine our processes. Over time, this results in higher consistency and reliability, decreasing the chance of errors that might compromise race outcomes,” says Gaudino.
To develop the pit cease evaluation answer, the mannequin was educated utilizing movies from the 2023 racing season and the YOLO v8 algorithm for object identification in SageMaker AI by way of the PyTorch framework. AWS Lambda and SageMaker AI are the core elements of the pit cease evaluation answer.
The workflow consists of the next steps:
- When a driver conducts a pit cease, entrance and rear movies of the cease are robotically pushed to Amazon Easy Storage Service (Amazon S3).
- From there, Amazon EventBridge invokes your entire course of by way of numerous Lambda capabilities, triggering video processing by way of a system of a number of Amazon Easy Queue Service (Amazon SQS) queues and Lambda capabilities that execute customized code to deal with vital enterprise logic.
- These Lambda capabilities retrieve the timestamp from movies, then merge the entrance and rear movies with the variety of video frames containing inexperienced lights to finally match the merged video with automotive and racing telemetry (for instance, screw gun habits).
The system additionally consists of using Amazon Elastic Container Service (Amazon ECS) with a number of microservices, together with one which integrates with its ML mannequin in SageMaker AI. Beforehand, to manually correlate the information, the method took a couple of minutes per pit cease. Now, your entire course of is accomplished in 60–90 seconds, producing close to real-time insights.
The next determine reveals the structure diagram of the answer.
Conclusion
The brand new pit cease evaluation answer permits for a fast and systematic evaluation of each pit cease to establish patterns and refine its processes. After 5 races within the 2025 season, Scuderia Ferrari HP recorded the quickest pit cease in every race, with a season greatest of two seconds flat in Saudi Arabia for Charles Leclerc. Diligent work coupled with the ML-powered answer extra effectively get drivers again on observe quicker, specializing in attaining one of the best finish end result attainable.
To study extra about constructing, coaching, and deploying ML fashions with totally managed infrastructure, see Getting began with Amazon SageMaker AI. For extra details about how Ferrari makes use of AWS providers, seek advice from the next further assets:
In regards to the authors
Alessio Ludovici is a Options Architect at AWS.
