Learn more about our autonomy capabilities with our CEO and founder, RJ Scaringe, and VP of Autonomy and AI James Philbin.

The Platform

The Rivian Autonomy Platform has been developed entirely in-house as a fully integrated system comprised of three core elements:

• A multimodal suite of 11 cameras and 5 radars offering sensor redundancy and a 360º worldview. Our in-house developed cameras have the highest resolution of any vehicle in North America, while the overlapping radars offer additional visibility, detecting objects over long-distances and through challenging conditions.
• An on-board compute module capable of over 200 trillion operations per second, making a Rivian vehicle the most powerful computer our customers likely own.
• Rivian-trained machine learning models based on the latest ML research and transformer architectures, which enable the vehicle to reason and understand its environment.

Understanding

Our vehicles use early sensor fusion—which combines raw information from the cameras, radars and other sensors into a unified world model. This is a more powerful and performant approach than late sensor fusion, which is more common in the automotive industry, that processes data from each sensor individually. We’re able to achieve this through our end-to-end approach to development and the powerful central compute module in every Rivian vehicle. With early fusion, we are able to deliver greater accuracy, reliability, and performance across a variety of challenging conditions, like dense fog, darkness, and intense sun glare.

Reasoning

The cohesive, fused model of the world around the vehicle is then processed in real-time using in-house machine learning algorithms running on our central compute module, giving Rivian vehicles a richer understanding of their surroundings. As a result, our vehicles are able to make better decisions based on what they sense—like finding a gap to execute a lane change, or smoothly reacting to movements from other vehicles. 

Machine Learning

Our onboard ML models are flexible by design and can accommodate different types of cameras and sensors, or positions of that hardware. This will allow the models to run on different vehicles with different sensor arrangements, enabling incredible headroom not only on each Rivian today, but also in future vehicles. A flexible design also allows us to move fast and respond to customer feedback, which will always be important to Rivian.

Off-board, we continuously distill and enhance the system through powerful, larger ML models fed by terabytes of new information from our vehicles on the road. All of this is accomplished while protecting our customers’ privacy through strict anonymization and aggregation protocols. 

What’s next

Our platform is continuously enhanced—nearly every month—through software updates. Later this year, we plan to expand our hands-off/eyes-on capabilities to more highways and to select surface roads. Next year, a hands-off/eyes-off feature is planned for controlled conditions with our current Gen 2 vehicles. 

This new era of autonomy is incredibly exciting. While being first to market with new capabilities is important, our #1 goal is being the most trusted. As our technology advances, our intention is to do it safely, responsibly, and ultimately make driving less stressful and to give you back your time. We can’t wait to show you what’s next.