Status: SHIPPING
AGIBOT X2
AGIBOTHumanoid robot
AGIBOT X2 is a humanoid robot line sold by AGIBOT (Shanghai) through its online store as the AGIBOT X2 Neo and the AGIBOT X2 Ultra (New Version), with an X2 EDU education variant and an older X2 Ultra hardware revision documented in AGIBOT's AimDK_X2 developer docs. The current flagship, X2 Ultra (New Version), is about 1.31 m tall, weighs about 39 kg and has 31 degrees of freedom, including 2 in the neck, 7 per arm, 3 in the waist and 6 per leg. It carries a LiDAR, an RGB-D camera, stereo and monocular RGB cameras and a head touch sensor, driven by an RK3588s+RK3588 motion-control computer and an NVIDIA Jetson Orin NX (157 TOPS) development computer. AGIBOT documents an AimDK_X2 secondary-development SDK built on ROS 2 with Python and C++ interfaces, a MuJoCo-based simulation environment, a reinforcement-learning deployment pipeline using ONNX Runtime, and open-source URDF/MJCF models on GitHub. Secondary development is documented for the Ultra-tier hardware (X2 Ultra, X2 Ultra (New Version), X2 EDU) but the main product page marks it 'Not supported' for the plain X2 unit.
AI-ASSISTEDOverview drafted from the cited sources with AI assistance. Awaiting editor review.
- HEIGHT
- ≈1.31 m
- WEIGHT
- ≈39 kg · ≈33 kg (X2 EDU) · ≈35 kg (X2 Neo)
- DOF
- 31 · 30 (X2 Ultra (URDF v1.3.0 flagship)) · 29 (X2 EDU) · 30 (X2 Neo)
- RUNTIME
- ≈2 h
Suggest an edit
Sources checked17 sources
DATA · 24 FIELDS
Specifications
- Official source
- Community measured
- Not published
| Height | ≈1.31 mX2 Ultra (New Version); identical dimensions given for X2 Ultra and X2 EDU. | Official source [6][9] |
|---|---|---|
| Weight | ≈39 kg · ≈33 kg (X2 EDU) · ≈35 kg (X2 Neo)X2 Ultra (New Version). | Official source [5][6][9] |
| Degrees of freedom | 31 · 30 (X2 Ultra (URDF v1.3.0 flagship)) · 29 (X2 EDU) · 30 (X2 Neo)X2 Ultra (New Version); total actuated DOF. | Official source [5][6][9] |
| Locomotion | Bipedal | Official source [3] |
| Walking speed | Up to 1.8 m/s (typical ≤0.8 m/s)Same figures published for X2 and X2 Ultra. | Official source [3][9] |
| Ingress protection | Not published | Not published |
| Peak joint torque | 120 N·mPublished as an overall peak figure; the specific joint is not named. Same value for X2, X2 Ultra, X2 Ultra(new version) and X2 EDU. | Official source [3][9] |
|---|---|---|
| Arm DOF (per arm) | 7 per armConsistent across X2 Ultra, X2 Ultra(new version), X2 EDU and X2 Neo. The main product page's older 'X2' SKU instead lists 5 DOF per arm (see discrepancies). | Official source [4][5][9] |
| Leg DOF (per leg) | 6 per legConsistent across X2 Ultra, X2 Ultra(new version) and X2 EDU. | Official source [9] |
| Waist DOF | 3Consistent across X2 Ultra, X2 Ultra(new version) and X2 EDU (yaw, pitch, roll per the main product page). | Official source [9] |
| Neck DOF | 2 · 1 (X2 Ultra (URDF v1.3.0 flagship)) · 0 (X2 EDU)X2 Ultra (New Version): adds a head-pitch axis versus the 1-DOF neck on the earlier X2 Ultra. | Official source [6][9] |
| Battery capacity | ≈421 WhAimDK_X2 developer docs figure, same for X2 Ultra, X2 Ultra(new version) and X2 EDU. The main product page instead states approx. 500 Wh (see discrepancies). | Official source [9] |
|---|---|---|
| Runtime | ≈2 hAt a continuous 0.5 m/s walking pace, not a general idle/standby runtime. | Official source [3][9] |
| Charging time | ≤1.5 h | Official source [9] |
| Compute | RK3588s+RK3588 + Jetson Orin NX (157 TOPS) | Official source [9][10] |
|---|---|---|
| Perception sensors | LiDAR, RGB-D, stereo/mono RGB, touch, IMU, GNSS (Ultra-tier only) | Official source [11] |
| Connectivity | Wi-Fi, Bluetooth, 4G/5G (+UWB on New Version) | Official source [9] |
| Arm reach (per arm, excl. end-effector) | 558 mmSame figure published for X2 Ultra, X2 Ultra(new version) and X2 EDU. | Official source [9] |
|---|---|---|
| End-effector | Optional OmniHand / OmniPicker | Official source [3][8] |
| Payload | 3 kg max / ≤1 kg full-rangeNot arm-specific; same figures for X2 Ultra, X2 Ultra(new version) and X2 EDU. | Official source [9] |
| Hand DOF | Not publishedOmniHand is documented as an optional module, but no DOF count for it is stated on the retrieved pages. | Not published |
| Indicative price | $25,600.00 USD (X2 Neo) · not published (X2 Ultra New Version)Store 'Sale price'; shipping ($500-$3000), duties/taxes and customs fees are additional per the same page. | Official source [5] |
|---|---|---|
| Availability | Orderable online (Neo) / contact sales (Ultra) | Official source [5][6] |
| Warranty | Not publishedAn 'AGIBOT X2 After-sales Service Guide' is listed in the Document Center navigation, but its content page was not retrieved/read. | Not published |
DATA · SOFTWARE & SIMULATION
Development
| INTERFACE | SUPPORT | PACKAGE / VERSION | SRC |
|---|---|---|---|
| SDK | Official | AimDK_X2: a task-level secondary-development framework with Python and C++ APIs, plus example code, an interface reference and a downloadable SDK package. Full secondary development is documented for X2 Ultra and X2 Ultra(new version); X2 EDU supports motion-control-only secondary development; the base 'X2' entry on the main product page is marked 'Not supported'. | [3][7][9] |
| ROS 2 | Official | AimDK_X2 is built on ROS 2 (Humble); ships a custom aimdk_msgs message package plus 'py_examples' and 'x2_rl_deploy_controller' ROS 2 nodes/topics for joint state, IMU and motion commands. | [7][13][14] |
| Isaac Sim / Lab | Not published | No Isaac Sim/Lab tutorial, importer or environment is documented for X2. The agibot_x2_urdf GitHub repository carries an 'issacsim' topic tag, but no Isaac-specific content, model or guide was found. | [16] |
| MuJoCo | Official | Official MJCF models (scene.xml, x2_ultra.xml / X2-Ultra.xml / X2-EDU.xml, plus omnihand/omnipicker variants) in the agibot_x2_urdf repository, and an official 'X2 MuJoCo Motion-Control Simulation' guide with local (ROS 2 Humble) or Docker deployment on Ubuntu 22.04. | [13][17] |
| Gazebo | Not published | No Gazebo model or integration is documented for X2. | |
| URDF / MJCF | Official | URDF and MJCF models for X2 Ultra (v1.3.0), X2 Ultra(new version) and X2 EDU (v1.4.0), including OmniHand/OmniPicker end-effector variants, open-sourced on GitHub (AgibotTech/agibot_x2_urdf) under the Mulan PSL v2 license. | [12][17] |
| Languages | Official | Python and C++ APIs, with example code for both in the AimDK_X2 documentation. | [7] |
| Teleoperation | Official | Handheld remote controller included as standard on X2 Ultra, X2 Ultra(new version) and X2 EDU; optional teleoperation accessories for X2 Ultra; AimDK docs cover remote-controller connection and PS5-gamepad control (ROS 2 'joy' node) during RL-policy testing. | [8][9][14] |
| Reinforcement learning deployment | Official | Official 'X2 RL Deploy' package: an ONNX-policy motion-control ROS 2 node (x2_rl_deploy_controller) paired with a MuJoCo simulation node and a joystick node; policies are trained externally and deployed via ONNX Runtime v1.23.2. | [14] |
TIMELINE · 2 EVENTS