The 20-year journey to fully autonomous cars with Dmitri Dolgov of Waymo

Dmitri Dolgov, co-CEO of Waymo, recounts his 20-year path in self-driving cars, starting as one of Google's first engineers on the project in 2009. Waymo now delivers over 500,000 fully autonomous rides weekly across 11 U.S. cities with about 3,000 vehicles, logging more than 4 million autonomous miles per week. Drawing from his Soviet-era upbringing and physics/math training in Russia and the U.S., Dolgov explains the technical foundations, evolution, and path to scaling.

Sensing and Real-Time Driving

A Waymo vehicle uses three complementary sensor types—LiDAR (high-resolution 3D mapping via laser pulses), radar (lower resolution but robust in fog, snow, or rain), and cameras (strong in clear conditions)—all providing 360-degree coverage. Microphones add audio cues, but the core trio feeds data into on-vehicle computers for real-time AI inference, with no cloud dependency for driving. Encoders process raw sensor data into world models (detecting objects like cars, pedestrians, roads, and signs), while a decoder generates actions like steering or braking via a vehicle interface. This setup handles multi-agent interactions, predicting behaviors in complex scenarios, such as nudging around a bus while detecting a hidden pedestrian via faint LiDAR reflections under it.