Rivian is making autonomous driving a central part of its technology strategy, developing its own processors, AI models, and sensor architecture rather than relying entirely on outside suppliers. Its immediate goal is Autonomy+, a point-to-point system expected on the R2 SUV around the end of 2026 and through updates to newer R1S and R1T vehicles, tells IEEE Spectrum. Autonomy+ remains Level 2+, meaning drivers must stay attentive.
Rivian’s system combines cameras, radar, and lidar. Data from these sensors feeds its Large Driving Model, an end-to-end neural network that converts raw information directly into steering, braking, and acceleration commands. Tracking objects across multiple frames helps the system distinguish persistent hazards from temporary sensor errors. The R2 will feature 11 high-definition cameras and five radars, with lidar planned to support future autonomy.
A major differentiator is Rivian’s decision to develop its own autonomy processor. The 5-nanometer RAP1 delivers 800 trillion operations per second. Two RAP1 chips in the Autonomy Compute Module provide 1,600 TOPS. Rivian says the module is eight times as powerful as its previous system while being 60% smaller. Its RivLink interconnect can combine additional processors as computing requirements increase.
Vehicle data is another important component. Rivian’s driving model currently draws cloud data from up to 125,000 vehicles. Unusual driving situations can be recorded, analyzed, and used in simulations to improve the model, with updated versions delivered through over-the-air updates.
Rivian ultimately wants to reach Level 4 autonomy. Backed by a $1.25 billion agreement with Uber, it plans to begin deploying robotaxis in 2028. Consumer vehicles could gain selected Level 4 functions by 2030. Yet technical progress alone will not determine success. Reliability, regulation, liability, and public trust remain major obstacles as Rivian competes with Tesla, Waymo, and established automakers for the future of autonomous transportation.
