How Rivian accelerated ADAS development on AWS (AUT318)

Rivian's ADAS Development on AWS

Key Takeaways

  1. ADAS (Advanced Driver Assistance Systems) has become a critical and strategic workload for automotive customers, with 90% of vehicles sold in the US having more than 5 ADAS features.
  2. Rivian has developed Polaris, a comprehensive data platform on AWS to accelerate their ADAS feature development.
  3. Polaris addresses key challenges around data ingest, toolchain complexity, and time-to-market for ADAS development.
  4. The platform leverages various AWS services and capabilities to enable efficient data management, simulation, and model training workflows.
  5. Rivian has been able to achieve significant milestones in their ADAS development by building Polaris on AWS.

Data Ingest and Management

  • Rivian uses AWS Data Transfer Terminals to accelerate data upload from test vehicles, improving data availability by 3x compared to traditional methods.
  • S3 is the primary storage service, leveraging features like intelligent-tiering and high-performance access through S3 Mountpoint.
  • Polaris includes a data discovery application that uses metadata to enable efficient search and access to the petabytes of collected data.

Toolchain Complexity

  • Rivian has built Polaris, a comprehensive platform with applications and workflows to manage the entire ADAS development lifecycle.
  • Key components include:
    • Autonomous Driving Data Framework (ADDF) for common tasks
    • Virtual Engineering Workbench (VEW) for pre-configured development environments

Accelerating Time-to-Market

  • Rivian leverages the latest GPU instances and capacity blocks on AWS to run large-scale simulations using AWS Batch.
    • Batch now supports multiple containers per job, simplifying the setup.
  • Polaris includes a SimDash application that enables efficient scenario management, dataset management, and job scheduling for simulations.
  • Rivian is exploring the use of generative AI for synthetic scene generation and augmentation to further accelerate their ADAS development.

Key Learnings

Rivian has shared several key learnings from their journey, including:

  • Compute optimizations, such as caching Docker images, monitoring GPUs, using good schedulers, and leveraging placement groups.
  • Storage optimizations, like enabling request metrics on S3, using intelligent-tiering, gp3 volumes, and leveraging the AWS Common Runtime.

Next Steps

  • Rivian plans to scale their data storage and compute capacity by 4x and 10x, respectively, to support faster feature delivery.
  • Exploring the use of generative AI for data curation and synthetic scene augmentation.
  • Optimizing inference costs for simulation and auto-labeling, and implementing stricter lifecycle policies on S3.
  • Closing the loop on data management by automatically creating datasets based on model performance and triggering new training jobs.

Your Digital Journey deserves a great story.

Build one with us.

Cookies Icon

These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.

If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.

Talk to us