Teleoperated 6DOF Robot Arms | Emiliano's Portfolio

LeRobot SO-100 Ubuntu Teleoperation 6-DOF Manipulator Motor Control Real-Time Control

A personal build of teleoperated SO-100 6-DOF robot arms using Hugging Face's LeRobot framework on Ubuntu. The system consists of a leader arm and a follower arm, where the follower mirrors the leader's joint positions in real time, enabling intuitive and responsive teleoperation for manipulation tasks.

The project involved assembling the SO-100 hardware, configuring the motor control stack, calibrating joint offsets, and integrating the LeRobot software pipeline for data collection and control. The system serves as a platform for collecting demonstration data for imitation learning and exploring dexterous manipulation research.

Setup & Demo

[ Robot arms photo / demo video placeholder ]

Technical Implementation

Software Stack

LeRobot Framework: Utilized Hugging Face's LeRobot for motor control, teleoperation, and data recording pipelines
Real-Time Teleoperation: Configured leader-follower control where the follower arm mirrors the leader's 6-DOF joint positions in real time
Ubuntu Environment: Set up and configured the full software stack on Ubuntu, including driver installation and serial communication
Data Collection: Established a pipeline for recording demonstration trajectories for future imitation learning experiments

Hardware

SO-100 Arm Setup

SO-100 Arms: Assembled two 6-DOF SO-100 robot arms with STS3215 servo motors
Motor Calibration: Calibrated joint offsets and limits for accurate position tracking between leader and follower
Serial Communication: Configured USB serial interfaces for low-latency motor command and feedback
Workspace Setup: Designed the physical workspace layout for effective teleoperation and task demonstrations

Results

Project Outcomes

Responsive Teleoperation: Achieved smooth, real-time leader-follower control across all 6 degrees of freedom
Successful Assembly: Completed full hardware assembly and software integration of both SO-100 arms
Data Pipeline: Established a working pipeline for recording manipulation demonstrations
Research Platform: Created a functional platform for exploring imitation learning and dexterous manipulation