Welding Robot

Autonomous weld-line tracking system for a UR10 robot arm. A camera detects the weld seam via ArUco markers and colour-based vision, and an operational-space controller follows the extracted trajectory while an admittance filter regulates contact force through a Schunk NetBox F/T sensor.

Overview

Component	Description
weld_line_vision_pkg	Detects workspace & weld line from a camera (DroidCam / USB), publishes nav_msgs/Path
cpp_controller	Operational-space controller (PD + admittance) that tracks the weld path in Cartesian space
tum_ics_ur10_robot	UR10 robot driver and bringup
tum_ics_schunk_netbox	Schunk NetBox force/torque sensor driver
ur_model_cpp	UR10 kinematics & dynamics model
utilities	Shared helper libraries

Prerequisites

• ROS Noetic on Ubuntu 20.04
• DroidCam installed on both iPhone and PC (used as the vision camera)
• Schunk NetBox F/T sensor connected via Ethernet
• UR10 robot reachable on the network

Building

cd ~/welding_robot
source /opt/ros/noetic/setup.bash
catkin build
source devel/setup.bash

Running the System

Open six terminals. In each one, first source the workspace:

source ~/welding_robot/devel/setup.bash

Terminal 1 — ROS Core

rosclean purge -y   # optional: clear old logs
roscore

Terminal 2 — UR10 Robot Driver

roslaunch tum_ics_ur_robot_manager robot_script_manager_ur10.launch

Terminal 3 — Force/Torque Sensor

roslaunch tum_ics_schunk_netbox sensor_publisher.launch

Terminal 4 — DroidCam (iPhone Camera)

Connect the iPhone via USB, open DroidCam on the phone, then:

droidcam-cli ios 4747 -v

Terminal 5 — Weld Line Vision + RViz

roslaunch weld_line_vision vision_with_robot_rviz.launch

This launches the vision pipeline (ArUco detection, weld-line extraction) and opens RViz for visualisation.

Terminal 6 — Controller

roslaunch cpp_controller operational_space_controller_6dof.launch

The controller moves through four phases:

1. INIT — Joint-space interpolation to a non-singular starting pose
2. LINEAR — Cartesian-space linear motion to the first weld-path waypoint
3. TRACK_SLOW — Follows the detected weld path with admittance-based force control
4. TRACK_FAST — Follows the detected weld path with admittance-based force control
5. END — Holds the final joint configuration

Key Topics

Topic	Type	Description
/weld_line_vision/weld_path	nav_msgs/Path	Detected weld trajectory (workspace frame)
/schunk_netbox/raw	geometry_msgs/WrenchStamped	Raw force/torque readings
/desired_trajectory_path	nav_msgs/Path	Controller's desired trajectory (RViz)
/admittance_state	Float32MultiArray	Admittance offset & external force (debug)

Further Documentation

• QUICK_START.md — Marker setup and vision quick-start guide
• VISUAL_CHECK_GUIDE.md — How to verify the vision pipeline
• force_torque_sensor_setup_guide.md — F/T sensor configuration
• gripper_setup_guide.md — Gripper setup instructions
• IMPLEMENTATION_SUMMARY.md — Detailed implementation notes