Skip to content
Robotoplus

Humanoid robots

Human-scale platforms, bipedal or wheeled, for manipulation and mobility in environments built for people.

6 products

The AGIBOT A2 Ultra humanoid robot seen head on and full length, standing with its arms at its sides, its white shell showing the dark grey joints at the shoulders, elbows and knees.

AGIBOT A2 Ultra

AGIBOT

Works at human height and human reach, so it uses the benches, shelves and controls a room already has instead of needing the space rebuilt around it. Its 40 joints and force sensing suit demonstration, service and research work rather than lifting: about 2 kg per arm at 1.2 m/s. The software stack is AimRT and open source, and the 14.4 Ah pack swaps for a charged one rather than stopping the shift.

View details
The AGIBOT X2 Basic humanoid robot standing upright on a plain white ground, front view, its white torso and limbs broken by black joint housings and yellow soles.

AGIBOT X2 Basic

AGIBOT

A compact humanoid at 131 cm and 33 kg, small enough to share a room with people and light enough for one person to reposition. It suits a reception, a stand or a teaching lab, where being approachable matters more than payload. The AimDK_X2 SDK ships with it in Python and C++, so a group can write its own behaviour rather than run a fixed demonstration.

View details
The AGIBOT X2 Ultra humanoid robot standing with its arms held slightly away from its body, front view, with yellow hands and yellow soles against a plain ground.

AGIBOT X2 Ultra

AGIBOT

The X2 that finds the part rather than being handed it: 3D LiDAR and RGB-D vision let it locate objects and move through a space nobody mapped by hand. It lifts 3 kg per arm, walks at 2 m/s and returns to its own docking station to charge, so nobody has to plug it in mid-shift. Two hours of runtime, 30 active joints, and the same AimDK_X2 SDK as the Basic.

View details
The AGIBOT G2 humanoid robot on its wheeled base, front view, the white upper body and arms rising from a dark rounded chassis instead of legs.

AGIBOT G2

AGIBOT

Built for work where the robot touches things with people nearby: all 26 joints are force-controlled, so contact is regulated rather than resisted. A wheeled base keeps it stable on flat industrial floors, its height adjusts between 1225 and 1795 mm to reach a low bench or a high shelf, and each arm carries 5 kg. Dual hot-swap batteries give about four hours. IP42 keeps out solid objects over 1 mm and dripping water; it is not a dust or washdown rating.

View details
The AGIBOT A3 humanoid robot standing upright on a white ground, front view, its pale shell showing the articulated joints at the hips, knees and shoulders.

AGIBOT A3

AGIBOT

The one to specify for shift work rather than a single demonstration. Batteries hot-swap in ten seconds, runtime reaches ten hours, and more than a hundred units can be coordinated as a group over UWB, so a fleet is a plan rather than an experiment. It walks at 1.8 m/s, runs at 5 m/s and carries 5 kg per arm with an actuator fitted. AGIBOT publishes 92 automotive-grade tests behind it and a design life of up to three years.

View details
The AGIBOT A3 Ultra humanoid robot standing upright, front view, its white shell and dark joints shown against a plain background.

AGIBOT A3 Ultra

AGIBOT

The A3 for work outdoors and off a mapped floor: all-terrain perceptive walking, GPS-RTK and UWB positioning, and teleoperation that keeps working beyond line of sight. It charges itself, navigates and does visual servoing without someone driving it, and its 51 joints end in dexterous hands rather than demonstration ones. Choose it over the standard A3 when the task leaves the building.

View details

Which humanoid, and why

The decision is usually made by the floor, not by the specification. A wheeled base like the G2 is stable and never has to solve balance, but it needs a flat, continuous surface and cannot manage a step. A bipedal platform like the A2 Ultra or the A3 handles thresholds, ramps and mixed terrain, and pays for it in runtime. All three publish a payload per arm: 5 kg for the G2 and the A3, about 2 kg for the A2 Ultra. Read them with their sources in mind. The G2's comes from AGIBOT's own parameter sheet, as does the A2 Ultra's; the A3's is a reseller figure and we hold no manufacturer table for that model, so it is the one to have confirmed before a job is planned around it.

Degrees of freedom is the number worth reading carefully.
The X2 Basic has 25 and no LiDAR or depth camera; the X2 Ultra has 30 plus 3D LiDAR, RGB-D and a docking station. That gap is the difference between a platform that performs a scripted interaction and one that can locate itself and a part without help. If the task involves finding something that moves, the sensing matters more than the joint count.
For manipulation, compare reach and torque rather than height.
The A3 publishes 320 Nm peak joint torque and 5 kg per arm with an actuator fitted; the A2 Ultra runs 40 active degrees of freedom with a six-DoF dexterous hand. Both are full-size, and they are aimed at different problems.
Runtime decides shift patterns.
The A3 publishes up to 10 hours with hot-swap packs; the A2 Ultra publishes roughly 3 hours standing and 1.5 hours or more walking. A platform that cannot cover a shift needs either a swap routine or a charging window designed into the process, and that is a layout decision, not a purchasing one.

If none of this resolves it, that is the normal outcome, and it is what a proof of concept is for: one task, on your floor, measured. We would rather tell you a robot is the wrong tool for a process than sell you one that proves it slowly.

What a first deployment involves

Choosing the platform is the short part. This is the shape of the work that follows, in the order it usually happens, because the second question after "which one" is "and then what".

The safety concept comes first, not last.
A machine of this size sharing a floor with people needs a risk assessment for your installation, a defined working envelope, a rule about who may enter it while the machine is active, and an agreed way to stop it. That assessment is part of the project and it needs a name against it on your side, usually whoever is already responsible for machine safety in that area. Starting it late is the single most common reason a deployment that works technically cannot be switched on.
Then the task is taught rather than described.
The application layer is built on ROS 2, and for most industrial work the reliable path is a taught sequence with vision used only where something genuinely moves. Where the task is too varied to teach directly, a person in a headset drives the machine through it while the system records, and those recordings become the training data. That is not a fallback, it is how this generation of machines is programmed.
Integration is where the schedule actually goes.
A humanoid that works in isolation and cannot be told what to do next by your existing systems is a very expensive island. Expect work on the interface to whatever holds the orders, whether that is a PLC, a warehouse system or a spreadsheet somebody maintains. This is ordinary integration engineering and it can be estimated, but it is rarely in the number a buyer has in their head.
Finally, the people.
Two or three operators who know how to start it, stop it, clear a fault and recognise when to call. A documented handover rather than a demonstration. Where nobody on site owns the machine, it quietly stops being used, and that has ended more deployments than any technical limit on this page.

Frequently asked questions

Contact us now