What is physical AI?AI with a physical body begins to work in the real world.
Physical AI is an area of AI that recognizes the real world and works by moving the body. Unlike AI, which generates text and images on a screen, it needs to produce results in a reality with physical laws and uncertainties. EmplifAI is working on social implementation in this area through the development of our own humanoid OpenShell.
Consult about physical AI development
Factory/warehouse transportation simulation
Unitree G1
In an Isaac Sim physics simulation, G1 held a 0.25 kg box with both hands, walked approximately 1.2 m, and placed it on another table. This is simulated footage.
RealMan RMC-AIDAL
We grasped the handle of a hypothetical transport case with a body size of 30 x 20 x 20 cm and a total weight of 10 kg, and verified the process of transporting it on wheels for approximately 1.2 m, placing it on a stand, and then releasing it. This is a physical simulation that corrects the center of gravity and contact settings of the public CAD model, and does not guarantee the load capacity of the robot.
RealMan RMC-AIDAL: 10 kg case transport with both arms
We verified the operation of holding a hypothetical transport case with a body size of 30 x 40 x 20 cm and a total weight of 10 kg with both hands using the left and right handles, transporting it approximately 1.2 m, and placing it on a stand. During transportation, contact between the left and right hands and load sharing are recorded. This is a physical simulation that corrects the center of gravity and contact settings of the public CAD model, and does not guarantee the load capacity of the robot.
You can also consult with us about simulations and demonstration experiments tailored to your business, such as replenishing shelves, supplying parts boxes, and sorting.
Environment/box 3D materials: NVIDIA SimReady / codepawl/simready-glb · CC BY 4.0 · Assets were converted, resized, and assigned adjusted surface materials.
3D digital twin with 10.8 million measured points that can be moved 360° on the browser
This is actual data measured while walking in EmplifAI's robot development lab (Unitree G1 humanoid/double-arm robot installation environment) for just 89 seconds using our own XGRIDS Lixel K2. In addition to being able to rotate and zoom 360° by dragging or swiping the mouse, you can experience switching between four display modes, ceiling slicing (cross-sectional cutting), and viewpoint jump to the robot installation spot on the spot.
What can be done with XGRIDS 3D spatial data × physical AI?
The greatest value of 3D scanning a site is not just in creating drawings for human viewing. By combining millimeter-level 3D spatial data with EmplifAI's humanoid robots (Unitree G1, etc.), dual-arm manipulators (OpenArm), VLA models, and the NVIDIA Isaac Sim environment, everything from pre-diagnosis for robot introduction to virtual training, autonomous movement, and remote control will change dramatically.
Digitize real space in 89 seconds and create a virtual verification environment where humanoid and dual-arm robots move
Even in factories, warehouses, and research facilities where drawings do not exist, you can obtain 3D point clouds and live-action textures of the entire room by simply walking around with XGRIDS Lixel K2. In the actual measurement data of the EmplifAI lab, the entire space (68.3 m2, 10.8 million points), including the humanoid by the window (Unitree G1) and the dual-arm robot work cell in the center, was converted into spatial data in just 89 seconds. By importing the acquired space into the NVIDIA Isaac Sim or ROS 2 environment, you can run simulations (Real2Sim2Real) of walking routes, whole body control, and grasping motions before bringing the robot to the site.
What can be achieved with this combination
- Build a full-scale 3D virtual space where humanoids and AMRs run and work by simply walking around the site.
- Collision detection model and navigation map conversion cooperation for NVIDIA Isaac Sim, USD, and ROS 2
- Environment generation for robot reinforcement learning and imitation learning that directly reflects on-site fixtures, workbenches, and obstacles
Watch the video: From an 89-second walkthrough to the creation of a 3D space with 10.8 million points
You can switch between the 360° multimodal analysis video of the lab space actually measured with XGRIDS Lixel K2 and the 10.8 million-point spatial build-up video created by laser scanning.
360° rotation × 3 modes continuous change (Full color RGB → Cyber elevation → LiDAR X-Ray)
This is a showcase video that rotates 10.8 million points of 3D spatial data generated from a single walking scan (89 seconds) and displays it in 360° rotation while seamlessly switching between three modes: live-action full color RGB, cyber elevation hologram by height, and X-Ray perspective based on laser reflection intensity.
"4K spatial analysis/drawing/inspection visualization" that can be developed from multiple angles from a single scan
From the acquired high-density 3D data to full-color point clouds, elevation holograms, 1m grid architectural ortho diagrams, CAD floor plans with automatic dimensions, millimeter heat maps of floor unevenness and wall verticality, and robot delivery clearance verification, you can develop analysis views that suit your purpose. Click on the thumbnail to switch, and click on the image to enlarge in high resolution.
Click to enlarge 4K high-definition image4-division multimodal 3D spatial analysis showcase
From one scan data, four modes can be simultaneously developed: (1) full-color RGB point cloud, (2) elevation heat map perspective, (3) 1m grid architectural orthoplan, and (4) cross-sectional profile. You can understand the appearance, dimensions, and structure of the space at a glance.
Meaning of physical AI
Physical AI is a term that refers to AI that recognizes the real world, makes decisions, and works through the body. It captures its surroundings with cameras and distance sensors, decides what action to take next, and moves its arms, wheels, and legs to affect objects and people. It may be easier to understand if you rephrase it as "AI with a body."
The reason why it has rapidly attracted attention in recent years is that large-scale models can now be applied to robot control. Until now, robot movements had to be programmed in detail by humans for each task. In physical AI, it is becoming more realistic to learn actions by showing people what they are doing, or to generate actions from verbal instructions.
However, this area is not as easy as generative AI. If you are generating sentences, even if you get some strange output, you can just rewrite it. The real body cannot be remade. It can break if you drop something on it, and it can be dangerous if you hit someone. The difficulty with physical AI lies in ensuring that it works in this "irreversible" situation.
Differences with generation AI
Even with the same AI, the required technology and evaluation methods will differ depending on the target.
The object to be dealt with is reality.
The input and output of generation AI is data. The output of physical AI is physical movement, which depends on conditions such as friction, weight, deflection, and slippage. The same instructions will not produce the same results every time.
The cost of failure is high
You can rewrite the text, but you can't get back what you dropped. Ensuring safety and reproducibility is as important, if not more important, than performance.
Not enough training data
Unlike the large amount of text and images available on the Internet, you have no choice but to collect robot movement data yourself. Data collection itself is a big part of development.
No delays allowed
Control has a fixed cycle. If the response is delayed, the posture will collapse and the movement will fail. Balancing the amount of calculation and response speed is a design constraint.
Evaluation depends on the field
The success rate in the actual field is more meaningful than the benchmark numbers. If the environment changes, such as lighting, floor surfaces, and people's movements, the results will change as well.
Inseparable from the design of the body
What you can learn changes depending on what kind of hand you use and how many degrees of freedom you have. Software cannot be developed in isolation.
Technology supporting physical AI
This is the elemental technology that EmplifAI is actually working on.
- imitation learning
- A method for learning robot movements from demonstrations of tasks performed by humans. Instead of writing down the movements in a program, we teach them through demonstration.
- VLA（Vision-Language-Action）
- A model that connects visual and linguistic understanding to action generation. Decide the next action based on verbal instructions and the situation in front of you.
- Sim2Real
- A method of transferring control learned and verified through simulation to the robot. A design that takes into account differences such as friction, mass, sensor error, and communication delay is required.
- Remote control/telepresence
- Technology that transfers an operator's physical movements and intentions to a robot in real time. It can be used both for actual operation before automation and for collecting learning data.
- Collection of training data
- The process of designing the target work, equipment, collection volume, and quality standards, and acquiring data using the robot. This will be the foundation for physical AI development.
- ROS 2 / LeRobot
- A foundation used to implement robot control and learning. Design including connection with existing robot SDK.
Application to humanoids
Humanoids are attracting attention as a potential source of physical AI. The reason is simple: you can use the environment created for people as is. The difference from dedicated machines is that it can be brought to the site without installing new dedicated equipment.
OpenShell, developed by EmplifAI, is a 28-axis humanoid that combines an expressive head, dual arms, and a trolley. I chose this configuration in order to study both the ability to work and how to interact with people at the same time. Having a face makes it easier for people to talk to you and conveys your intentions. Although this does not appear in the performance table, it is definitely effective in the workplace where people are present.
When implementing something, it is more realistic to start with remote control and gradually expand the scope of automation, rather than aiming for complete automation from the beginning. Learning data is accumulated as the site is operated remotely, and this data becomes the material for automation.
Support for physical AI development
OpenShell
A humanoid developed in-house. The specifications as a research and development platform are listed.
Sim2Real development support
Development support from simulation to robot verification using NVIDIA Isaac Sim/Isaac Lab.
Robot learning data collection contract
We undertake the collection of data necessary for imitation learning and operation verification, from condition design to delivery.
Robot learning data acquisition kit
We will propose the configuration of the data acquisition environment, including remote control equipment, cameras, and recording PCs.
Robot SI/Introduction support
We provide consistent support from the selection of AI robots to their introduction and operation.
Achievements/case studies
We are posting initiatives such as imitation learning, VLA implementation, and remote/body-sharing telepresence.
NVIDIA GEAR Sonic × Unitree G1: Real robot demonstration
We ran NVIDIA GEAR Sonic on a Unitree G1 robot and verified its operation. Watch the video to see the demonstration.
Gemini 4 Argon × Unitree G1: Turning Vision into Action
Unitree G1 just picked up a plastic bottle using Gemini 4 Argon. No hard-coded motions. Just AI turning vision into action. The future of robotics is getting real.
Even when the bottle is moved and placed again, the AI visually understands the scene and generates the grasping motion. We provide Physical AI and VLA implementation on the Unitree G1 as well as custom PoC and verification development tailored to your operations.
Frequently asked questions
What is the difference between physical AI and robotics?
Robotics refers to robotics in general, including mechanisms, controls, and sensors. Physical AI is a term that emphasizes the role of AI in recognizing the real world and determining actions. The two are not contradictory; physical AI is built on robotics.
Can physical AI be used in practice right now?
It depends on the purpose. Although there are still many issues to be solved before completely replacing human work, it can be put into practical use at this point in applications where "interacting with people" itself is valuable, such as customer service, guidance, and attracting customers, and in operations that involve remote control.
Can you make robots learn your work?
It depends on the target work. First, organize your work and see if you can obtain demonstration data and define evaluation metrics. It is safe to start with a PoC focused on one task. If you let us know the target work and current situation, we will suggest how to proceed.
How much training data do I need?
It depends on the difficulty of the task, the degree of variation in the environment, and the desired success rate. We first conduct test collection, determine the required quantity and quality standards, and then proceed to actual collection.
Which robot should I start with?
For research and development, it will be easier to use models and SDKs that are compatible with secondary development. We will propose robot that are actually used in verification, such as Unitree G1 and OpenShell. You can also rent it and try it out before deciding.
Consultation on physical AI development
Please let us know the target robot, the work you want to accomplish, and your current development environment. You can consult with us even when requirements and technology selection have not been decided. It is also possible to start with a PoC for one task.
Consult about physical AI development