An independent field manual for physical AIRev. 2026.11 · 7 chapters · 7 articles
Physical AI Handbook
Physical AI · Robotics · Safety · Deployment

The field manual for robots that leave the lab.

How robot models work, what makes a deployment safe, which standards apply, and what it takes to run machines around people — explained plainly, with sources.

Start with the foundations →Browse all articles
Diagram of a robot arm surrounded by three zones — protective stop, reduced speed and warning — with a person standing outside them. PROTECTIVE STOPREDUCED SPEEDWARNINGS — SEPARATION DISTANCE
Fig. 1 — Speed and separation monitoring: the robot slows, then stops, as a person approaches.

§Contents

All chapters →
  1. 01

    Foundations

    What physical AI is, how the stack fits together, and the vocabulary you need.

    1 article
  2. 02

    Robot Models

    Vision-language-action models, world models, and the policies that turn pixels into motion.

    1 article
  3. 03

    Safety & Risk

    Risk assessment, functional safety, and what changes when the controller is a neural network.

    1 article
  4. 04

    Deployment

    Getting robots out of the pilot and into production: sites, integration, operations, cost.

    1 article
  5. 05

    Hardware

    Actuators, hands, sensors, compute and batteries — the body the intelligence has to live in.

    1 article
  6. 06

    Simulation & Data

    Where robot training data comes from: teleoperation, simulation, video, and the sim-to-real gap.

    1 article
  7. 07

    Standards & Regulation

    ISO 10218, the EU Machinery Regulation, the AI Act, and who is liable when something goes wrong.

    1 article

§Latest

All articles →
Start here · 01 Foundations

What is physical AI? A working definition

Physical AI is AI that perceives the real world and acts on it through a machine body. What the term covers, how the stack works, and why it is hard.

14 Nov 20263 min read

In brief
  • Physical AI is AI that perceives the physical world, decides what to do, and acts through a machine body — a robot arm, a mobile robot, a humanoid or a vehicle.
  • It differs from generative AI in one decisive way: its output is motion with physical consequences, not text or pixels that a person reviews first.
  • A physical AI system is a stack — sensors, a learned model, low-level control, hardware and an independent safety layer — and it is only as reliable as the weakest of them.

§Glossary

All 21 terms →

Physical AI

AI systems that perceive the physical world, decide what to do, and act on it through a machine body such as a robot arm, a mobile robot, a humanoid or a vehicle.

Vision-language-action model

A model that takes camera images and a natural-language instruction as input and outputs robot actions. Most are built by adding action output to a pretrained vision-language model.

Sim-to-real gap

The drop in performance when a policy trained in simulation runs on a physical robot, caused by differences between simulated and real physics, sensing and timing.

Speed and separation monitoring

A collaborative safeguarding method in which sensors track the distance between a person and the robot, and the robot slows and then stops before that distance falls below a calculated protective separation distance.

Functional safety

The part of a machine’s safety that depends on a control system performing correctly — for example, stopping the robot when a light curtain is broken. Its required reliability is expressed as a performance level or safety integrity level.

Teleoperation

A human controlling a robot remotely in real time. It is used both to collect demonstration data for training and to recover robots in the field when autonomy fails.