OpenArm Technical Documentation
1. System Overview
OpenArm is a fully open-source, anthropomorphic robotic arm platform with 7 degrees of freedom (7-DOF) per arm, designed for physical AI research, contact-rich manipulation, and data-centric robotics workflows.
The system exposes the entire hardware–software stack, including mechanical design, actuation, firmware, control interfaces, simulation assets, and manufacturing documentation. This enables researchers and developers to build, modify, analyze, and deploy real robotic systems without reliance on proprietary components.
OpenArm is intended for research, experimentation, and system development, rather than consumer or safety-certified industrial deployment.
2. Design Goals and Principles
The design of OpenArm is guided by four primary goals:
2.1 Accessibility
Provide research-grade humanoid manipulation capabilities at a cost significantly lower than traditional humanoid or industrial robotic arms.
2.2 Full Openness
All core system components are open-source, including:
Mechanical CAD files
Electrical schematics
Firmware and low-level motor control
High-level control software
Simulation models (MuJoCo / Isaac Sim)
Complete bill of materials (BOM) and assembly guides
2.3 Real-World Practicality
The platform is designed for continuous operation, repeated experimentation, and real-world interaction, including contact-rich tasks and data collection.
2.4 Safety-Oriented Interaction
Mechanical compliance, backdrivability, joint limits, and controlled inertia are prioritized to support safe human–robot interaction in controlled environments.
3. Mechanical Architecture
3.1 Degrees of Freedom and Kinematics
7 DOF per arm, enabling human-like redundancy and dexterity
Joint layout supports complex manipulation, orientation control, and reachability
Designed for both single-arm and bimanual configurations
The kinematic structure closely follows human arm proportions to simplify teleoperation, imitation learning, and intuitive control.
3.2 Anthropomorphic Proportions
OpenArm adopts dimensions comparable to a human arm corresponding to an approximate body height of 160–165 cm, providing:
Effective workspace coverage
Manageable inertia for safe interaction
Motion profiles suitable for contact-rich tasks
Stable dynamics for learning-based control
Key physical parameters per arm include:
Reach: 633 mm
Weight: 5.5 kg
3.3 Structural Design and Mounting
Aluminum alloy structural frame
Stainless steel base plate (8 mm thickness)
Uniform M6 threaded mounting grid
Designed for secure desktop or fixture mounting
Each joint incorporates mechanical hard stops to prevent overextension and ensure safe motion boundaries.
4. Actuation System
4.1 Motor Selection Strategy
OpenArm employs DAMIAO-series integrated joint motors, selected per joint based on load, speed, and control requirements:
Joint Location
Motor Model
Design Rationale
Shoulder joints
DM-J8009P-2EC
High torque for primary load-bearing
Elbow joints
DM4340 series
Balance of torque, size, and rigidity
Wrist joints
DM-J4310-2EC
High-speed response and dexterity
Gripper joint
DM-J4310-2EC
Precise force and motion control
The architecture combines QDD (Quasi-Direct Drive) characteristics with compact geared joints to balance:
Backdrivability
Structural rigidity
Compact form factor
4.2 Payload Definitions
Payload capacity is defined under standardized test conditions:
Rated payload: 4.1 kg
Sustained hold for 1 minute at full extension
Peak payload: 6.0 kg
Dynamic lift-and-hold test over a short-duration motion
⚠️ Payload values include the mass of the end effector.
For example, with a 1.5 kg end effector installed, usable payload is reduced accordingly.
5. End Effector System
5.1 Standard Gripper
Maximum opening width: 88 mm
Direct-drive motor actuation
Linear motion guided by bearings and sliders
Designed for force feedback and bidirectional control
The gripper mechanism supports smooth motion, precise grasping, and contact-aware manipulation.
5.2 Custom End Effector Interface
OpenArm supports easy customization of end effectors via replacement of the J8_B interface component, allowing:
Custom grippers
Sensorized tools
Experimental manipulation devices
Design guidelines emphasize:
Mechanical alignment and load balance
Cable routing and strain relief
Manufacturability via CNC or 3D printing
6. Electrical and Control Interfaces
6.1 Power and Communication
Operating voltage: 24V DC
Control bus: CAN / CAN-FD
Control frequency: up to 1 kHz
The high-frequency control loop enables low-latency torque, velocity, and position control, essential for learning-based and force-controlled applications.
6.2 Low-Level Control API
The low-level API provides:
Direct joint-level access
Position, velocity, and torque control modes
Real-time feedback with minimal latency
This layer is designed for:
System identification
Advanced control research
Integration with custom controllers
7. Software Ecosystem
7.1 Robot Description and Modeling
Modular XACRO files
Generated URDF models for simulation and control
Compatibility with major robotics frameworks
7.2 ROS2 Integration
OpenArm provides native ROS2 support, including:
Standardized ROS2 nodes and message interfaces
RViz visualization
Compatibility with MoveIt and motion planning tools
7.3 Simulation Support
OpenArm is fully supported in:
MuJoCo
Accurate contact dynamics
Suitable for reinforcement learning
Isaac Sim
High-fidelity physics and rendering
Sim-to-real validation
Simulation models are calibrated to closely match physical dynamics, supporting large-scale parallel training and transfer learning.
8. Control and Teleoperation
OpenArm supports advanced control modes, including:
Real-time gravity compensation
Impedance and force control
Bilateral teleoperation with force feedback
Smooth, human-like motion profiles
These capabilities make OpenArm suitable for human demonstration capture, teleoperated manipulation, and contact-intensive tasks.
9. Data-Centric System Design
Beyond manipulation, OpenArm is designed as a robotic data generation platform:
Structured recording of real-world trajectories
Support for imitation learning datasets
Alignment between simulated and physical data
Repeatable, high-quality demonstrations
This enables workflows spanning:
Simulation
Physical execution
Learning
Evaluation
10. Application Domains
10.1 Research
Imitation learning and skill acquisition
Reinforcement learning and sim-to-real transfer
Human–robot interaction
Contact-rich manipulation
10.2 Education
University robotics laboratories
Embodied AI training programs
Robotics competitions and coursework
10.3 Industrial Exploration
Light assembly and inspection
Vision-guided manipulation
Human–robot collaboration research
11. Safety and Operational Guidelines
OpenArm incorporates multiple safety mechanisms:
Mechanical joint limits
Backdrivable joint design
Emergency stop support (hardware and software)
Clear installation and operational procedures
The system is intended for trained users in controlled environments and is not certified for safety-critical or industrial autonomous deployment.
12. Manufacturing and Availability
OpenArm is available in two configurations:
DIY Kit — full component set for self-assembly
Fully Assembled System — calibrated and ready for deployment
CEREBOTO is the officially recommended manufacturing partner, offering:
Verified hardware consistency
Optimized supply chain
Localized support and faster delivery
13. Project Status and Contributions
OpenArm is an actively developed open-source project.
Contributions from researchers, developers, and industry partners are encouraged.
The objective is to provide a transparent, extensible, and realistic humanoid manipulation platform for the next generation of physical AI research.