PhysDBPhysical AI Map

Control

Robot policy

The mapping from robot observations and goals to actions, often learned from demonstrations, reinforcement learning, or model outputs.

policyactions

What it is

A policy is the executable decision layer. It may be end-to-end, modular, or hybrid with planners and controllers.

Why it matters

Most Physical AI claims eventually have to pass through a policy that moves hardware.

How not to overread it

Policy behavior is embodiment- and environment-dependent.

Related edges

supports training

LeRobot

Policy training

Reproducibility still depends on hardware and data.

feeds

State estimation

Policy inputs

Bad state estimates can make good policies fail.

tests

Manipulation

Object interaction

Narrow task success is not broad reliability.

executes or constrains

Control stack

Hardware commands

Model output must respect control constraints.

runs

Onboard compute

Onboard inference

Compute spec does not imply task success.