Robot arms

NexArm ROS

A desktop ROS 2 robot arm with 3D vision, voice and an OpenClaw agent, on Raspberry Pi 5 or NVIDIA Jetson.

Raspberry Pi
NVIDIA Jetson
Suggested for university

NexArm ROS is a desktop robot arm that Hiwonder built for advanced ROS education and embodied AI research. It pairs a Raspberry Pi 5 or NVIDIA Jetson controller with an ESP32 controller: the first handles vision, voice and AI, the second drives the servos and monitors power. Magnetic encoder bus servos move the arm, a parallel-rail gripper does the grasping, and a 3D depth camera on the end of the arm lets it recognise, track and pick up objects in three dimensions. An AI voice module and multimodal large AI models let it act on spoken instructions, and the OpenClaw agent breaks larger tasks into steps. It runs ROS 2 Humble, Hiwonder supplies tutorials, ROS and ESP32 source code and system images, and the same arm is the base of Hiwonder's NexArm ROS sandbox, sliding rail, dual-arm conveyor and mobile chassis kits.

Zeta42 is an authorised Hiwonder partner in the Middle East.

Configurations

Choose how it ships.

Model: Without Raspberry Pi Controller · With Raspberry Pi 5 (4GB) · With Raspberry Pi 5 (8GB) · With Raspberry Pi 5 (16GB) · Without Jetson Controller · With Jetson Nano(4GB) · With Jetson Orin Nano Super(4GB) · With Jetson Orin Nano Super(8GB) · With Jetson Orin NX Super(8GB) · With Jetson Orin NX Super(16GB)

What it does

Built for.

3D vision and grasping
The structured light depth camera fuses colour and depth (RGB+D), so NexArm can detect and track objects, recognise faces, measure an object's distance, height and volume, and pick it up in 3D space. Hiwonder's inverse kinematics lets the arm reach any point in its workspace.
Voice and large AI models
The AI voice box listens and speaks, and large language, speech and vision models let NexArm act on spoken requests: sorting waste into bins, tracking a named object, describing what is in front of it or moving blocks one after another.
OpenClaw agent
OpenClaw breaks a high-level instruction into steps such as recognition and reporting, remembers earlier commands and scenes, and can be reached through Telegram, the PC, the app or by voice.
Two controllers
A Raspberry Pi 5 or NVIDIA Jetson runs ROS 2, vision and AI, while the ESP32 controller handles servo control, feedback and power monitoring. Hiwonder recommends a Jetson Orin controller for deeper development of AI models; the Jetson Nano and Raspberry Pi 5 run the default algorithms. Kits without a controller come with the matching expansion board and enclosure, but no microSD card or SSD.
Smooth, steady motion
Dual-shaft magnetic encoder bus servos report temperature, voltage and position, and trapezoidal acceleration and deceleration keep movement smooth. The industrial bearing base supports 270° of joint rotation and reduces wobble, and an OLED screen shows the arm's status.
ROS 2 tools and software
NexArm supports Gazebo simulation, MoveIt 2 motion planning and YOLO26 model training. The PC software has action group, coordinate and AI modes with joint tuning and live feedback, and the arm can also be driven from the app or a wireless controller.

Where it is used

Applications.

University courses in ROS 2, inverse kinematics and machine vision
Embodied AI research that links large AI models and agents to a physical arm
Training and deploying YOLO object detection models
Sorting, palletising and pick-and-place exercises in 3D

Videos

See it run.

Specifications

The full sheet.

Dimensions: 245 × 131 × 617 mm
Material: Metal brackets
ROS controller: Raspberry Pi 5 / Jetson Nano / Jetson Orin Nano / Jetson Orin NX
ESP32 controller: ESP-WROOM main chip
ESP32 controller detail: ESP-WROOM + AT32F421 dual-chip board, 0.96 in OLED status display
Camera: 3D structured light depth camera
Servo models: HX-12H / HX-30HM / HX-65HM intelligent bus servos
Servo stall torque: HX-12H: 12 kg·cm at 11.1V; HX-30HM: 30 kg·cm at 11.1V; HX-65HM: 65 kg·cm at 12.6V
Payload: 500 g at a 0.5 m operating range
Repeatability: ±2 mm
Base: Industrial bearing base, 270° joint rotation
Control methods: ROS topics/nodes, Python programming, app control, PC software control
Power supply: 12 V 5 A DC power adapter
Operating system: Ubuntu 22.04 + ROS 2 Humble
Software: iOS/Android app, Windows PC software
Communication: USB, Wi-Fi, Ethernet
Programming tools: Python, C, C++
Storage: 64 GB TF card (Raspberry Pi 5 / Jetson Nano); 128 GB SSD (Orin Nano / Orin NX)
Included materials: Tutorials, ROS source code, ESP32 source code, system images, supporting software