Waymo and TU Delft Develop Reference Driver Model Mimicking Human Collision-Avoidance Behavior

U.S. autonomous driving company Waymo has partnered with Delft University of Technology in the Netherlands to develop a computational cognitive model called ReD (Reference Driver), designed to simulate the full process of human drivers' perception, decision-making, and control actions in near-collision scenarios. The model aims to enhance the safety performance of autonomous driving systems by addressing limitations in current testing methods, which typically focus on specific scenarios or single metrics and struggle to fully replicate the entire sequence—from hazard detection to actual evasive maneuvers. Human drivers rely on the coordination between their central and peripheral nervous systems to execute collision-avoidance reactions almost instantaneously, and the ReD model seeks to reproduce this complex behavioral mechanism within autonomous driving systems.