Autonomy Strategy: Rivian Acquires Third Dimension to Digitize Fleet Video
The 15-person UK firm will use generative AI to reconstruct rare driving scenarios from 400,000 daily clips for self-driving validation.
By Thiago Kovalenko ·
Rivian has acquired Third Dimension, a 15-person software company based in the United Kingdom, to convert daily fleet footage into three-dimensional testing environments. The acquisition targets the specific gap in autonomous vehicle training where rare, high-risk road events occur too infrequently in real life to provide sufficient data for machine learning models.
TechCrunch first reported the deal, which Rivian confirmed without disclosing a purchase price. Third Dimension was founded in 2024 by Tolga Kart, a former Tesla Autopilot simulation engineer, and Piotr Sokólski, who spent four years building neural simulators at Wayve. The company developed generative AI tools designed to create photorealistic 3D environments compatible with game engines, originally targeting clients in film, gaming, and the defense sector.
The entire Third Dimension team is now integrated into Rivian's autonomy and AI group under James Philbin. Philbin stated that simulation is central to training and validating the company's driving systems. He noted that the acquired technology transforms real-world data into detailed 3D environments, allowing engineers to isolate and analyze specific variables without waiting for physical occurrences.
The data source for these simulations consists of video clips sent by owners of the R1S, R1T, R2, and commercial vans who opt into camera data sharing. Philbin said the fleet transmits more than 400,000 driving sequences daily. While most of this footage captures routine driving, the scenarios that challenge autonomous systems are rare, such as a mattress in the fast lane or a child running between parked cars. By digitizing these events, Rivian engineers can recreate them in 3D and alter conditions to test system responses.
This acquisition places Rivian in a crowded field of simulation efforts. Waymo unveiled its Waymo World Model in February, leveraging Google DeepMind's Genie 3 technology to generate unencountered scenarios. Tesla has also demonstrated a neural world simulator for its Full Self-Driving software. The move by Rivian expands its engineering presence in London, where most of the Third Dimension staff are based.
The deal arrives as Rivian faces tight deadlines tied to its partnership with Uber. In March, Uber agreed to invest up to $1.25 billion in Rivian to deploy R2 robotaxis, with $300 million paid upfront. The remaining funds are contingent on autonomy milestones, including placing 10,000 driverless R2 vehicles on Uber's network in San Francisco and Miami starting in 2028.
Consumer timelines have also shifted. Rivian's point-to-point driving feature is scheduled to reach a limited group of invited Autonomy+ owners later this year, subject to nondisclosure agreements. CEO RJ Scaringe originally promised "eyes-off" driving in 2026 but has moved that target to 2027. Level 4 capability is planned for 2028 on the Gen 3 hardware unveiled at the company's AI & Autonomy Day.
