Rivian Autonomy+ is being readied for public roads, with a test drive in Palo Alto showing the system guiding an R1S through traffic lights, crosswalks, and stop signs while the route flows past Silicon Valley landmarks. The company says owners will be able to enter an address and let the vehicle handle point-to-point driving to any mapped destination in the United States and Canada.
The feature is pitched as one of the most capable semiautonomous systems headed to market, and a stepping-stone toward future robotaxis and, ultimately, broader consumer self-driving capability. After years of slow progress, autonomy has become a top priority for automakers and investors amid the wider surge in artificial intelligence.
Rivian plans to launch the point-to-point experience on its new R2 SUV around the end of this year, with over-the-air updates bringing it to the latest R1S and R1T. Pricing is set at $49.99 per month, or a $2,500 upfront option.
Tesla’s competing Full Self-Driving (Supervised) subscription is $99 per month, and Mercedes intends to charge $3,950 for a three-year MB.Drive Assist Pro plan on the CLA-Class EV, which still requires at least one hand on the wheel.
Despite its capabilities, Rivian Autonomy+ is categorized as Level 2+, meaning drivers must remain attentive and ready to take control. Rivian and a host of rivals are pushing toward higher levels.
Level 3 would allow drivers to disengage briefly in limited conditions, while the industry race centers on Level 4, where vehicles can operate without a human driver inside defined areas and conditions.
Robotaxis in select cities in the United States, China, and the Middle East have already demonstrated driverless Level 4 operation, though at limited scale. Rivian’s approach uses a blend of cameras, radar, and lidar, paired with in-house silicon to run an AI autonomy model.
Backed by a $1.25 billion agreement with Uber, Rivian targets a Level 4 robotaxi fleet launch in 2028, beginning in San Francisco and Miami and expanding to additional North American and European cities. Data gathered from commercial and consumer fleets is expected to accelerate the learning cycle for future consumer features.
Rivian Autonomy+ strategy and timeline
Rivian remains a small player by volume, selling about 42,000 vehicles last year across the R1S, R1T, and Electric Delivery Van. Tesla delivered roughly 1.6 million vehicles and Toyota more than 11 million.
To compete, Rivian is leaning on software and electronics. Volkswagen agreed to invest up to $5.8 billion in a joint venture, Rivian and Volkswagen Group Technologies, to access Rivian’s electrical architecture and software for the R2, which went on sale in June.
The joint venture does not include Rivian’s autonomous technology. Separately, Rivian’s R2 platform underpins its $1.25 billion deal to supply Uber with up to 50,000 robotaxis, with an initial 10,000 slated to begin service in 2028 in San Francisco and Miami before expanding to 25 cities across the United States, Canada, and Europe.
Rivian acknowledges reliability challenges and high repair costs on earlier models and says the R2 addresses those issues. The competitive landscape remains intense.
- Tesla: reports 1.1 million active users of its FSD system.
- Toyota’s Woven by Toyota: is collaborating with Waymo on autonomy platforms for robotaxis and consumer cars.
- Elon Musk: has said Tesla’s Level 4 capability for general consumers could arrive as early as the fourth quarter of 2026, after prior delays.
Rivian chief executive RJ Scaringe has said some Level 4-derived functions, such as self-parking, could begin appearing in showroom vehicles by 2030. In the nearer term, the company is focused on scaling Level 2+ and preparing for Level 3 features.
How the system learns to drive with Rivian Autonomy+
Rivian fuses camera, radar, and lidar feeds into a single pipeline interpreted by a neural network the company calls its Large Driving Model. The end-to-end model processes raw sensor inputs and outputs vehicle controls for steering, braking, and acceleration.
It identifies and tracks objects over time, favoring persistent detections across multiple frames to avoid reacting to transient sensor noise. Commands are executed via redundant by-wire systems for steering and braking.
In a demonstration of Rivian Autonomy+, a company driver remained behind the wheel while an autonomy engineer observed. The system kept to posted limits, smoothed stops at lights, and handled speed bumps conservatively.
Waymo has reported significantly lower serious crash rates than human drivers across its autonomous miles, though broader consumer deployment will test how higher levels of autonomy perform outside controlled fleets.
Rivian says its core model can ingest cloud data from up to 125,000 vehicles for analysis and validation before updates are sent to customer cars via monthly over-the-air releases. Owners must opt in to data collection, and onboard systems prioritize recording unusual scenarios.
Rivian is also reducing reliance on high-definition maps and continuous connectivity by emphasizing onboard perception and localization, improving resilience in tunnels or dense urban areas.
The R2 will use 11 high-definition cameras and five radars, with a compact lidar unit integrating into the roofline early next year to support future autonomy. Executives say lidar costs have fallen from more than $10,000 to a few hundred dollars in under a decade.
Each sensor type adds complementary strengths. Cameras excel at color and classification but can struggle in low light.
- Cameras: excel at color and classification but can struggle in low light.
- Lidar: provides reliable 3D geometry and performs in darkness and glare.
- Radar: penetrates rain and snow though at lower spatial resolution.
Why Rivian built its own chip
To process the data flood, Rivian designed the Rivian Autonomy Processor, or RAP1, a 5-nanometer chip capable of 800 TOPS. Two RAP1 chips in the Autonomy Compute Module deliver a combined 1,600 TOPS and about 5 billion pixels per second throughput.
The company says this surpasses the pixel throughput of Nvidia’s Drive AGX Thor, which delivers 1,000 TOPS per chip and about 3.5 billion pixels per second. Taiwan Semiconductor Manufacturing Co. is slated to fabricate RAP1 to Rivian’s specifications.
Rivian argues that co-developing software and hardware cut a year from its program and enabled optimization exclusively for autonomy rather than sharing compute with infotainment or other functions. The latest vehicle architecture consolidates control units into zonal domains and adopts early fusion, aligning raw sensor data in time and space before neural processing.
This approach preserves richer information but requires significant compute, which RAP1 is meant to provide. The system is designed to scale, with RivLink enabling additional modules for higher autonomy levels.
Rivian’s road to Level 3 and beyond
The company’s next milestone is Level 3 on highways, enabling hands-off and eyes-off driving in limited conditions. Rivian has not provided a launch date.
Critics warn Level 3 can erode driver vigilance, and several automakers are confining initial implementations to controlled freeway scenarios. Rivian says if a driver fails to respond to prompts, the car will slow, pull to the shoulder, or contact emergency services.
Regulatory hurdles and public trust
Adoption of partially to fully automated features is projected to grow from 8 percent of new vehicles in 2024 to 28 percent by 2030 in developed markets, according to Morgan Stanley. Analysts argue the wider benefits will materialize when advanced systems migrate from robotaxi fleets to retail cars, freeing drivers from commuting chores and enabling new services.
Yet significant barriers remain. Public skepticism is high, some jurisdictions are reassessing pilot programs, and liability could shift from drivers to manufacturers as systems progress from Level 2 to Level 3.
Labor groups are wary of job losses, and a patchwork of state and local rules complicates deployment. High-profile mishaps, even without injuries, have dented confidence.
Experts say the path forward will depend as much on transparency and regulation as on technical prowess. The industry is being urged to share safety data, engage with oversight, and build trust before scaling.
Updated on 08 September 2026.