Solutions › Guides › Unitree A2 payload guide

Unitree A2 payload guide: mounting, power and sensors that fit

Short answer: The Unitree A2 (about 820 × 440 × 570 mm, 37 kg, OEM reference payload about 25 kg) has plenty of carrying margin, so the real limits are where you mount things, how you power them and what you block. Start from a back plate or shell, keep the front and rear LiDAR fields of view clear, and budget power for compute first; a full inspection build with a thermal gimbal usually stays well under a third of the rated payload.
Updated 2026-11-19 · YUANZHI ROBOTICS engineering team

The Unitree A2 is a mid-size industrial quadruped of about 820 × 440 × 570 mm and 37 kg, with a reference payload of about 25 kg and its own front and rear LiDAR plus HD cameras. Weight capacity is rarely the problem on this robot. What decides whether a payload works is the mounting interface, the power and data path, and whether your sensors leave the robot's own sensors and legs free to do their job.

Choose the mounting base first

Every A2 build starts with something that bolts to the back and carries everything else. These are the base options in our configurator, with their own mass so you can count it in the budget:

BaseMassBest for
Quick-release rail plate (10 mm aluminium, dual rails)0.65 kgResearch and prototyping, sensors that change often
Carbon-fibre shell0.95 kgWeight-sensitive builds, long patrols
Integrated aluminium shell with compute bay1.6 kgInspection robots that run every day; compute and wiring hidden inside
Arm-reinforced shell with flange2.1 kgMounting a light manipulator
Double-deck shell with extra battery space2.4 kgPayloads that need their own battery for long shifts

A rail plate is the fastest way to get sensors on the robot, but cables stay exposed and every bracket is a separate part. A shell costs more design time and pays it back in weather resistance, clean cable routing and a lower centre of gravity.

Unitree A2 with a YUANZHI ROBOTICS form-fitting sensor shell, camera, antennas and lights (3D render)
Unitree A2 with a YUANZHI ROBOTICS form-fitting sensor shell, camera, antennas and lights (3D render)

Keep the native sensors and legs clear

The A2 relies on its front and rear LiDAR and cameras for terrain perception. Anything mounted at the front or rear ends of the back must stay out of their field of view, including antennas, cable loops and gimbal arms at their extreme pan and tilt angles. Check three things on a 3D model before cutting metal:

Sensors that fit well on an A2

SensorExampleMassPower
Mapping LiDARLivox Mid-360265 g6.5 W
Mapping LiDAR, 16-channelHesai PandarXT16800 g9 W
Depth cameraIntel RealSense D435i75 g2.5 W
Thermal dual-spectrum mini gimbalHikvision DS-2TD5566T-15/V24.5 kg18 W
Four-gas module (CO/H2S/O2/CH4)Winsen ZCE04B50 g0.5 W
Dual-antenna RTK moduleUnicore UM9821.8 g0.6 W
Fixed acoustic imagerFluke SV600850 gnot published

For choosing between thermal gimbals see best thermal gimbal for a robot dog, and for LiDAR mounting and mapping see LiDAR SLAM payloads for robot dogs.

Power and compute

Compute is usually the biggest electrical load on the back. A Jetson Orin NX 16GB module is rated up to 40 W and handles LiDAR-inertial mapping, navigation and a few detection models. A Jetson AGX Orin 64GB is rated up to 60 W and is worth it when you run several cameras with heavy models or reconstruction on board. Add sensors on top and plan the power tree: a DC-DC stage with enough headroom, separate fused branches for compute and sensors, and a clean shutdown path so the computer is not cut off mid-write when the robot powers down.

Whether to draw from the robot battery or carry a second battery depends on shift length and on the power output the OEM exposes on your A2 version, so check the current Unitree documentation before you fix the design. The trade-offs are covered in robot payload power distribution.

A worked example: daily inspection build

ItemMassPower
Integrated shell with compute bay1.6 kg3 W
Hikvision DS-2TD5566T-15/V2 thermal gimbal4.5 kg18 W
Livox Mid-3600.265 kg6.5 W
RealSense D435i0.075 kg2.5 W
Winsen ZCE04B gas module0.05 kg0.5 W
Jetson Orin NX 16GBwith carrier, not countedup to 40 W
Total listedabout 6.5 kgabout 70 W peak

That is roughly a quarter of the reference payload before compute carrier, cables, comms and antennas, so there is room for a 5G module, lights or a speaker. Peak power matters more than average here: size the converter for all sensors plus compute at full load.

Robot dog payload shell with gimbal, antennas and lights (3D render)
Robot dog payload shell with gimbal, antennas and lights (3D render)

Software around the payload

Hardware is half the job. An A2 inspection payload normally needs drivers for every sensor, time synchronisation between LiDAR, cameras and IMU, a mapping and localisation stack that does not depend on the robot's own black-box navigation, and a mission layer that sends the robot to inspection points and stores results. If you already have a robot, see who can develop software for a robot dog.

Next step

Pick the base and sensors in the payload configurator to get weight, power and compute totals for your own build, and see the Unitree A2 platform page. YUANZHI ROBOTICS designs and integrates robot payloads, builds inspection robots and handles quadruped secondary development.

FAQ

How much payload can the Unitree A2 carry?

Unitree gives a reference payload of about 25 kg for the A2. For inspection work you rarely need that much; a thermal gimbal, LiDAR, depth camera, gas module and mounting shell come to roughly 6.5 kg before compute and cabling, which leaves margin for runtime and stair climbing.

Can I add my own LiDAR to the Unitree A2 if it already has front and rear LiDAR?

Yes. The native LiDARs serve the robot's own locomotion and obstacle sensing. A separate Livox Mid-360 (265 g, 6.5 W) or Hesai PandarXT16 (800 g, 9 W) is often added on top for mapping and your own localisation stack, mounted so it does not block the native sensors.

What compute should I put on a Unitree A2 payload?

A Jetson Orin NX 16GB (up to 40 W module power) covers mapping, navigation and a few detection models. Choose a Jetson AGX Orin 64GB (up to 60 W) if you run several cameras with heavy models or 3D reconstruction on board.

Who can design and integrate a payload for the Unitree A2?

YUANZHI ROBOTICS designs form-fitting shells, brackets, harnesses and power distribution for the A2, and develops the mapping, navigation, recognition and ground-station software around them.

Configure your payload and get a quote

Get a quoteWhatsApp