Waymo, a leader in autonomous vehicles, has developed a new computer model that simulates the internal 'surprise' a human driver feels just before a crash. This advanced simulation, known as Reference Driver (ReD), aims to set a higher standard for Waymo robotaxi performance benchmark validation in 2026. It delves into human cognitive states, moving beyond simple reactive maneuvers to understand complex pre-collision scenarios.
Autonomous vehicles are engineered to remove human error, but Waymo invests heavily in modeling complex human cognitive errors and reactions to prove their safety. This reveals a critical insight: simply removing human drivers is insufficient. AVs must anticipate and respond to the consequences of human error, or even simulate human-like decision-making for robust safety validation.
This human-centric simulation will likely accelerate public trust and regulatory acceptance of autonomous vehicles. It provides a more robust, relatable safety benchmark. Developed with TU Delft, the model compares Waymo's software against human behavior, according to TechCrunch. This collaboration validates AV safety against nuanced human decision-making.
How Waymo's ReD Model Simulates Human Cognition
The Reference Driver (ReD) model simulates human driver behavior in the run-up to a crash, including the internal 'surprise' a driver feels, according to TechCrunch and The Verge. Unlike previous models focused on last-second maneuvers, ReD models cognitive states and pre-collision behavior. This moves Waymo's robotaxi performance benchmark beyond simple collision avoidance, integrating a deeper understanding of human cognitive responses.
ReD analyzes not just crash avoidance, but how the autonomous system's response compares to a human driver's cognitive process before an incident. This detailed simulation offers new insight into robotaxi performance in complex, real-world conditions, challenging traditional metrics focused only on immediate outcomes. The implication: true AV safety demands understanding the 'why' behind human errors, not just avoiding their physical consequences.
