Product Description

NexArm is an open-source AI robotic arm platform built for embodied AI and VLA (Vision-Language-Action) learning. Combining high-precision motion control, AI vision, large AI model interaction, and the LeRobot ecosystem, it covers the entire development workflow from data collection and imitation learning training to real-world deployment.
Featuring a K230 AI vision module, NexArm handles voice interaction, visual perception, and hand-eye coordination. It also seamlessly integrates with mobile chassis, sliding rails, and conveyor belts to build full-scenario embodied AI systems.

High-Precision Magnetic Encoder Bus Servo

Built around a dual-output-shaft magnetic encoder servo system with all-metal gears. A powerful magnetic motor and real-time feedback from multiple algorithms ensure precise and stable joint movement.

Advanced Inverse Kinematics Algorithms

NexArm can detect target objects within the recognition area, calculate their position coordinates and placement angle, and adaptively adjust each joint through inverse kinematics to enable flexible grasping.

Built for Imitation Learning

Supports a leader-follower teleoperation system composed of a leader arm and a follower arm, with deep compatibility at the underlying level with the latest Hugging Face LeRobot framework.

K230-Powered Hand-Eye Coordination

Powered by the K230 module with 6 TOPS of computing power, NexArm supports 30+ AI vision functions including YOLO, as well as multimodal large-model deployment for advanced AI applications in complex scenarios.

Color Recognition

NexArm can accurately identify objects of different colors within its field of view, with different preset actions triggered through programming.

Face Recognition

Backed by the K230 chip’s 6 TOPS of equivalent compute, NexArm can recognize and remember different faces, then trigger different programmed responses.

Gesture Recognition

Using the K230 vision module’s YOLO algorithm, NexArm can accurately recognize different hand gestures and respond with different actions.

AI Waste Sorting

Powered by deep learning on the K230 vision module, NexArm can recognize different waste cards and sort them to designated locations based on instructions.

VisualObject Tracking

With large vision model analysis, NexArm can track and lock onto any object within its field of view.

Complex Task Management

NexArm can break down complex voice commands on its own and place items in the corresponding areas to complete complex handling tasks.

Smart HomeAssistant

NexArm can send the live scene in view to a large vision model and combine it with instructions to generate matching language responses and actions.

Intelligent Sorting

With support from a large vision model, NexArm can identify target objects in view, adjust its posture in real time based on instructions, and pick and sort them into designated locations.

Cross-Platform Control and Connectivity

App Control

Through the app, each NexArm joint can be controlled independently. Designated objects can also be grasped and sorted with flexible control, while preset action groups can be called up quickly.

Offline Drag Teaching

The leader arm can be dragged offline for intuitive demonstration. Combined with advanced imitation learning algorithms, the follower arm can accurately reproduce motion and grasping trajectories.

Wireless Multi-Device Control

Supports multilingual remote collaboration across multiple devices, with unicast, multicast, and broadcast modes. Automated scenarios can be deployed with ease, helping build a highly responsive AIoT collaborative ecosystem.

Wireless Controller Control

Equipped with a PS3 wireless controller and low-latency one-to-one pairing, allowing multiple devices to run side by side without interference. Programming support also enables different feedback actions on NexArm.

High-Performance Hardware

Specification Parameters