Daily Car
·01/01/2026
Rivian recently offered a select group of journalists a preview of its upcoming point-to-point autonomous driving system. This advanced driver-assistance technology aims to rival Tesla's Full Self-Driving capabilities, handling complex urban and highway driving scenarios. The demonstration provided an early look at Rivian's approach to creating a more generalized and adaptable self-driving experience.
During a roughly 20-minute test ride in an R1S, the vehicle demonstrated its ability to navigate various driving situations smoothly. It came to stops behind other vehicles, respected red lights, and executed turns and lane changes competently. While the experience was largely uneventful, there were minor instances where the system exhibited hesitation, requiring brief manual acceleration adjustments from the driver.
Rivian's "Large Driving Model" (LDM) operates similarly to large language models, learning from extensive real-world driving data. Unlike older systems that relied on explicit rules, Rivian's LDM is trained end-to-end, absorbing data and generating driving outputs. This approach aims for greater generalizability, allowing the system to handle situations like speed bumps or traffic lights without explicit programming for each scenario. The model learns these behaviors implicitly from the training data.
Training an autonomous system on human driving data presents challenges, as the data includes imperfect driving habits. Rivian has worked to train out tendencies like excessive acceleration on open roads or rolling through stop signs. The company plans to further enhance its system with the introduction of lidar-equipped R2 crossovers in late 2026, which will improve perception and data quality for the LDM. Rivian also has long-term goals for eyes-off driving and eventually Level 4 autonomy.









