Material Insights · Understanding the Material World

Autonomous driving: Science

Principles

The science it rests on, with why each principle matters to it

Relies on
  • Artificial neural networks Perception relies on deep neural networks, which learn to recognise objects from examples rather than from written rules [7].
  • State estimation (Kalman filter) Kalman-type filters merge noisy readings from many sensors into one best estimate of the car's position and of other road users [6].
  • Feedback control theory Steering and speed controllers compare where the car is with where it should be and correct the error continuously.

as of 2026-11-11

Discoveries

The discoveries that made it possible: when, by whom and what was found

YearWhoWhat
1958Frank RosenblattThe perceptron, an early learning machine, shown to the public [7]
1960Rudolf KalmanA filter that estimates a system's state from noisy measurements, published in March 1960 [6]
2012AlexNet teamA deep network trained on GPUs and large data sets starts today's approach to machine vision [7]

as of 2026-11-11

The science explained

The science behind it explained for a curious reader

Picture a careful learner driver who never blinks. Every fraction of a second the car asks three questions: where am I, what is around me, and what will it do next? For the first, it blends a map, satellite positioning and its own motion sensors, trusting each according to how noisy it is, which is the idea behind the Kalman filter used on Apollo 11 [6]. For the second, neural networks trained on millions of examples label what the cameras and lidar see [7]. For the third, it predicts the paths of others and plans its own, then steers and brakes to follow the plan, checking and correcting as it goes. The hard part is the rare, odd event, such as a person in fancy dress or debris on the road.

as of 2026-11-11

Sources

  1. Honoring a legacy algorithm Aerospace America (AIAA), 2016
  2. CHM Releases AlexNet Source Code Computer History Museum (Hansen Hsu), 2025