Rivian acquires UK unit for 3D street tests

More than 400,000 driving sequences from Rivian’s customer fleet return to the company every day. Rivian has acquired Third Dimension, a 15-person UK startup, to convert that data into 3D test environments for its self-driving system. TechCrunch first reported the deal, which Rivian did not price.

Third Dimension was founded in 2024 by Tolga Kart and Piotr Sokólski. Kart previously spent more than two years on Autopilot simulation at Tesla and served as a senior director on Call of Duty at Activision. He later led engineering at Parallel Domain. Sokólski spent four years at Wayve building photorealistic neural simulators for self-driving vehicles.

The company raised a $7 million seed round led by Felicis in October 2024. It described itself as a generative AI tool that creates 3D environments for game engines. Potential customers included game studios, film production, the US military, and autonomous vehicle companies.

All staff will join Rivian’s autonomy and AI group, led by James Philbin. The acquisition expands Rivian’s presence in London, where most of the team is based. Philbin told TechCrunch that simulation is central to training and validating autonomous driving systems. He added that Third Dimension has built technology that turns real-world data into detailed 3D environments.

The system uses footage from owners of the R1S, R1T, R2, and commercial vans who opt into camera data sharing. While most daily footage is uneventful, edge cases that challenge autonomous systems are rare. Examples include a mattress in the fast lane, a police officer waving traffic through a red light, or a child running between parked cars.

Engineers can recreate these real scenes in 3D and alter variables rather than waiting months for such events to occur in the wild. This approach allows for rapid testing of rare and complex driving scenarios.

Competitors are pursuing similar strategies. Waymo unveiled its World Model in February, built on Google DeepMind’s Genie 3, to generate scenarios its vehicles have not encountered. Tesla has also demonstrated a neural world simulator for its Full Self-Driving software.

The acquisition occurs amid time pressure from Uber, which agreed in March to invest up to $1.25 billion in Rivian to deploy R2 robotaxis. Only $300 million is paid up front. The remainder depends on autonomy milestones, starting with 10,000 driverless R2 vehicles on Uber’s network in San Francisco and Miami from 2028.

Rivian’s consumer timelines have shifted. Point-to-point driving for invited Autonomy+ owners is due later this year, with testers signing nondisclosure agreements. CEO RJ Scaringe moved the target for "eyes-off" driving from 2026 to 2027. Level 4 capability is planned for 2028 on Gen 3 hardware.

Simulation serves as a training and testing tool rather than proof of safety. A system may perform well on generated edge cases yet still fail on a real street. Rebuilding real scenes from real data provides a close approximation to actual road conditions for validation purposes.