How gjcliff/frankateleop works
Project type and purpose
• This is a full-stack robotics teleoperation system built on ROS2 Iron.
• Its goal is to let a user control an Emika Franka Panda arm in real or simulated hardware via hand-tracking gestures using MediaPipe and MoveIt Servo.
File structure
Top-level you see five ROS2 packages plus workspace files:
• handcv — Python node for camera input, MediaPipe hand tracking & gesture recognition
• cv_franka_bridge — Python node that subscribes to handcv output and translates it into motion commands
• franka_teleop — C++ node using MoveIt Servo to drive the robot via geometry increments
• numsr_franka_moveit_config — MoveIt configuration files (URDF/xacro, controllers, planning params, RViz setups)
• hand_interfaces — custom ROS2 msgs and srvs definitions used by the other nodes
Each package has its own launch files, configs, resource files (meshes, RViz), and tests.
README for additional insights
• The README clarifies the data flow: RealSense Camera → handcv → cv_franka_bridge → franka_teleop → Franka Robot
• It explains external dependencies (MediaPipe, RealSense-ROS, MoveIt) and how to run on real vs. simulated hardware.
• Gestures (thumbs up/down, open/close fist) map to start/stop, shutdown, gripper open/close.
Elements
• Clear labels: node names, topic/service names, package names
• Blue for perception nodes
• Green for bridging
• Orange for control
• Gray for hardware
• Yellow for configuration files/params
• External dependencies: MediaPipe library, MoveIt, RealSense driver