Intelligent robotic control system combining LeRobot framework with ChatGPT-4o Vision API for autonomous manipulation of a LeKiwi mobile manipulator with manual teleoperation, vision-based control, and SLAM navigation modes.
This robotics integration combines the LeRobot framework from HuggingFace with ChatGPT-4o Vision API for intelligent control of a LeKiwi mobile manipulator. The system enables three operational modes including manual teleoperation via leader-follower arm control, vision-only autonomous control for object manipulation, and SLAM-enabled autonomous navigation with real-time mapping capabilities.
The platform features a 6-DOF robotic arm with gripper mounted on a 3-wheel omnidirectional mobile base, dual-camera system with front and wrist-mounted cameras for environmental awareness, RPLidar A1 for 2D SLAM functionality, and real-time ChatGPT-4o Vision integration for AI-driven decision making. The system runs on a Jetson Orin Nano 8GB controller with ROS2 Humble for advanced navigation features.
Communication flows through a distributed architecture where user commands are sent from a laptop controller to the Jetson via ZMQ messaging at 30Hz for motor control, while camera images stream back via HTTP REST endpoints. The system successfully demonstrates autonomous pick-and-place tasks with multiple test objects, showcasing full ChatGPT reasoning during task execution.
Powerful features that make this solution stand out
Manual teleoperation via leader-follower arm control mirroring movements at 30Hz, vision-based autonomous manipulation using AI without mapping, and full autonomous navigation with laser-based SLAM for position-aware operations.
6-DOF robotic arm with gripper on 3-wheel omnidirectional base providing full workspace coverage, precise object manipulation capabilities, and seamless integration with vision and navigation systems.
Front-mounted 640x480 USB camera for environmental awareness and 8MP wrist camera for detailed manipulation views, enabling comprehensive scene understanding for AI-driven autonomous operations.
Real-time ChatGPT-4o Vision API integration analyzing camera feeds to generate intelligent motor commands for object manipulation tasks with full reasoning transparency during execution.
RPLidar A1 laser scanner with ROS2 Humble SLAM packages enabling real-time 2D mapping and localization for position-aware autonomous navigation in dynamic environments.
Jetson Orin Nano for robot control with ZMQ messaging at 30Hz, laptop controller for AI processing, Flask HTTP endpoints for camera streaming, and multi-device network coordination.
Get a customized quote for your business needs
Enquiring about: LeRobot — AI Robotics