Fundamentals
Humanoid Robot & Physical AI Glossary (Key Terms)
HRS TeamUpdated 3 min read
Quick answer
This glossary defines the key terms used in humanoid robotics and physical AI in plain English — grouped into core concepts, AI and learning, hardware, and deployment and commercial terms. Use it as a quick reference when reading about humanoid robots, from actuators and degrees of freedom to vision-language-action models, sim-to-real and robot-as-a-service.
Humanoid robotics borrows vocabulary from mechanical engineering, AI and manufacturing. Here are the terms you will meet most often, grouped by theme.
Core concepts
| Term | Plain-English definition |
|---|---|
| Humanoid robot | A robot built in the general shape of the human body so it can work in spaces and with tools made for people. |
| Cobot (collaborative robot) | A robot arm designed to work safely beside people without fencing, using force and speed limits. |
| Industrial robot | A fast, powerful robot arm for high-volume fixed tasks, usually kept behind safety fencing. |
| AMR | Autonomous mobile robot — a wheeled robot that navigates and moves goods, but has no arms. |
| AGV | Automated guided vehicle — a wheeled transport robot that follows fixed routes or markers. |
| Degrees of freedom (DoF) | The number of independent joints a robot can move; more DoF means more flexible movement. |
| Bipedal | Walking on two legs, as opposed to using a wheeled or tracked base. |
AI and learning
| Term | Plain-English definition |
|---|---|
| Physical AI | AI that perceives and acts in the real world through a body, such as a robot. Also called embodied AI. |
| VLA model | Vision-language-action model — AI that turns camera images and a plain-language instruction into robot movements. |
| Foundation model | A large, general AI model trained on broad data that can be adapted to many tasks. |
| Imitation learning | Teaching a robot by showing it demonstrations of a task, often via teleoperation. |
| Reinforcement learning | Training a robot to improve through trial, error and feedback signals. |
| Teleoperation | A human remotely controlling a robot — used to operate it and to gather training data. |
| Sim-to-real | Training a robot in simulation, then transferring what it learned to the physical robot. |
| Digital twin | A realistic virtual copy of a robot or workspace, used for training and testing. |
Hardware and the body
| Term | Plain-English definition |
|---|---|
| Actuator | The motor at a joint that creates movement; humanoids have dozens. |
| End effector | The tool at the end of a robot arm — a gripper, hand or other device that does the work. |
| Gripper | An end effector that grasps objects, from simple two-finger types to multi-fingered hands. |
| Payload | The maximum weight a robot can carry or manipulate. |
| IMU | Inertial measurement unit — a sensor that tracks motion and orientation, essential for balance. |
| Proprioception | A robot's sense of where its own joints and limbs are positioned. |
| Dexterity | How skilfully a robot can manipulate objects with its hands or end effector. |
Deployment and commercial
| Term | Plain-English definition |
|---|---|
| Brownfield | An existing facility built for people, deployed into with minimal change — versus a purpose-built greenfield site. |
| Machine tending | Loading, unloading and monitoring a machine such as a CNC or moulding press. |
| Cycle time | The time a robot takes to complete one repetition of a task. |
| Uptime / availability | The share of scheduled time a robot is actually working and productive. |
| RaaS | Robot-as-a-service — subscribing to a robot for a recurring fee instead of buying it outright. |
| TCO | Total cost of ownership — all costs of running a robot over time, not just its purchase price. |
| ROI | Return on investment — the value a robot creates measured against what it costs. |
| Integrator | A partner that selects, deploys, integrates and supports robots for a specific operation. |
Frequently asked questions
- What does VLA stand for in robotics?
- VLA stands for vision-language-action. A VLA model takes in camera images (vision) and a plain-language instruction (language) and outputs the movements a robot should make (action) — the AI engine behind modern general-purpose robots.
- What is the difference between an AMR and a humanoid robot?
- An AMR (autonomous mobile robot) is a wheeled robot that transports goods but has no arms. A humanoid robot has a full body with hands, so it can manipulate objects and use spaces and tools built for people.
Continue learning
- What Is a Humanoid Robot? A Plain-English DefinitionA humanoid robot is built in the shape of the human body so it can work in spaces and with tools made for people. How they work and what they do.
- How Do Humanoid Robots Work?Humanoid robots sense their surroundings, decide with onboard AI, and move precise electric joints to act. Inside the full sense–think–act loop.
- What Are Vision-Language-Action (VLA) Models?A vision-language-action (VLA) model turns camera images and a plain-language instruction into robot movements. How VLAs work and why they matter.
- Humanoid Robots vs. AMRs and AGVsHumanoid robots vs AMRs and AGVs: wheels versus hands, what each does best, and how they work together in a facility.
See a humanoid robot work your task
HRS helps UK manufacturers select high-fit tasks, run real factory trials and prove ROI — with full integration, safety and long-term support.