General-purpose humanoid robots: Science
Principles
The science it rests on, with why each principle matters to it
Relies on- Feedback control theory Walking on two legs needs constant measurement and correction; Honda's first walking robot was built to study the principle of two-legged walking [3].
- Artificial neural networks Learned skills come from neural network models, such as a vision-language-action model that turns sight and instructions into movement [9].
as of 2026-11-11
Discoveries
The discoveries that made it possible: when, by whom and what was found
| Year | Who | What |
|---|---|---|
| 1868 | James Clerk Maxwell | Analysed the dynamics of the governor, a feedback device [11] |
| 1943 | Warren McCulloch and Walter Pitts | A mathematical model of the neuron, the basis of artificial neural networks [10] |
| 1986 | David Rumelhart, Geoffrey Hinton and Ronald Williams | Showed how networks with hidden layers learn by backpropagation [10] |
as of 2026-11-11
The science explained
The science behind it explained for a curious reader
Standing on two legs is like balancing a broom on your palm. The broom always starts to fall, and you keep it up by moving your hand under it. A humanoid robot does the same: sensors measure its tilt and motors shift its feet and hips to stay underneath its centre of mass. This is feedback control, which Maxwell analysed mathematically in 1868 [11]. Balance is only the start. To fold a box or pick a part from a bin, the robot needs skills that are hard to write as rules. Modern humanoids learn them from data with artificial neural networks, an idea traced back to a neuron model of 1943 and made practical by learning methods of the 1980s [10].
as of 2026-11-11
Sources
- History of robotics development Honda Global Corporate Website
- Gemini Robotics brings AI into the physical world Google DeepMind, 2025
- Scientific Background to the Nobel Prize in Physics 2024 Royal Swedish Academy of Sciences, 2024
- A History of Automatic Control C. C. Bissell (chapter in a Springer handbook, ISBN 978-3-540-78830-0; Lund University course copy)