SO-ARM 101
Building a 6-DOF robotic arm from scratch: assembly, calibration, and intelligent control
Overview
The SO-ARM 101 is a six-degree-of-freedom robotic arm based on the Arctos Robotics open-source design. This project documents the full journey, from unboxing and assembly through calibration, programming, and eventually intelligent autonomous control.
The arm uses STS3215 servos and a custom control board, with the goal of building toward real-world manipulation tasks using inverse kinematics and machine learning.
Project Journey
Research & Planning
Evaluated open-source robotic arm platforms. Selected the SO-ARM 101 based on the Arctos Robotics standard for its community support, 6-DOF capability, and extensibility.
Parts & Assembly
Received the full kit: 3D-printed structural components, STS3215 servos, wiring harness, and control board. Began mechanical assembly following the Arctos documentation.
Calibration & First Movements
Calibrated servo ranges and zero positions. Achieved first coordinated joint movements through direct serial commands.
Inverse Kinematics & ML Integration
Implementing IK solvers for end-effector positioning and exploring imitation learning for pick-and-place tasks.
Gallery
Key Features
6 Degrees of Freedom
Full spatial manipulation with shoulder, elbow, and wrist articulation
STS3215 Smart Servos
Position, speed, and torque feedback from each joint for precise control
Open-Source Design
Built on the Arctos Robotics standard, fully modifiable and community-supported
Real-Time Control
Low-latency serial communication for responsive joint coordination
ML-Ready Architecture
Designed to integrate imitation learning and reinforcement learning pipelines
Python & ROS2 Compatible
Scriptable via Python with a path toward ROS2 integration for advanced robotics
Technology Stack
STS3215 Servos
High-torque smart serial bus servos with position and temperature feedback
Arctos Control Board
Custom microcontroller board for coordinating 6-axis servo commands
Python
Primary scripting language for control logic and ML experiments
NVIDIA Isaac Sim
Physics simulation environment for training and testing in virtual scenarios
3D Printing (PLA/PETG)
Structural components printed for lightweight, customizable assembly