Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/MRRP-lab/arm-demos/llms.txt

Use this file to discover all available pages before exploring further.

Running arm-demos on physical hardware requires three components to be connected and configured before you launch: the UR3e arm over a local Ethernet network, a vacuum gripper wired to the robot’s digital I/O, and an OAK-D Pro camera attached via USB. Each component has its own setup step, and none of them depend on each other, so you can work through them in parallel if you have the hardware available.
Always keep an emergency stop (e-stop) accessible and within reach whenever the UR3e is powered and the driver is running. The arm will execute planned trajectories without further confirmation.

​
UR3e robot arm

The robot driver connects to the UR3e over a LAN interface. The default robot IP expected by the launch-lan task is 192.168.56.101.

​
Network configuration

Configure the network interface on your workstation with a static IP on the same subnet as the robot (for example, 192.168.56.1/24). On the robot’s teach pendant, navigate to Settings → System → Network and set:
FieldValue
IP address192.168.56.101
Subnet mask255.255.255.0
Default gateway192.168.56.1
Verify connectivity before launching: 
ping 192.168.56.101

​
External control URCap

The UR robot driver requires the External Control URCap to be installed and active on the pendant. Load the external_control.urp program on the pendant and start it before running pixi run launch (or pixi run run). The driver will wait for the URCap to connect before accepting motion commands.
The launch-lan task sets headless_mode:=true and launch_dashboard_client:=false. The driver communicates over the robot’s primary interface port (30001) and RTDE port (30004) — make sure these are not blocked by a firewall.

​
Vacuum gripper

The vacuum gripper is controlled by toggling digital output pin 0 on the UR3e controller using the ur_msgs/srv/SetIO service. The VacuumControl node in pick_and_place.py calls set_digital_out(0, ON) to engage suction and set_digital_out(0, OFF) to release. Wire the vacuum solenoid valve so that it is driven by DO0 on the robot’s tool I/O connector. Confirm the gripper actuates correctly with a manual test before running pick-and-place demos.

​
OAK-D Pro camera

The OAK-D Pro is mounted at the wrist and connects to the host via USB 3. The depthai library communicates with the camera directly over USB, so no additional ROS driver package is required.

​
udev rule (Linux)

By default, only root can access the OAK USB device. Add a udev rule so your user can open it without sudo: 
echo 'SUBSYSTEM=="usb", ATTRS{idVendor}=="03e7", MODE="0666"' | \
  sudo tee /etc/udev/rules.d/80-movidius.rules && \
  sudo udevadm control --reload-rules && \
  sudo udevadm trigger
You can also run this through Pixi: 
pixi run --environment oak-viewer setup-oak
Replug the camera after applying the rule to confirm it is visible to your user account.

​
WSL2 users

USB devices are not automatically forwarded to WSL2. Use usbipd on the Windows host to bind and attach the OAK camera. The connect-oak-wsl task handles this automatically, polling for VID 03E7 and re-attaching whenever the camera’s PID changes (which happens during firmware boot): 
pixi run --environment oak-viewer connect-oak-wsl
This launches a PowerShell script with elevated privileges. Leave it running in a separate terminal for the duration of your session.
Install usbipd-win on the Windows host if it is not already present. The script requires usbipd to be on the Windows PATH.

​
MediaPipe hand landmarker model

The gesture-control demo requires a pre-trained MediaPipe hand landmarker model that is not included in the repository. Download it with: 
pixi run download-mediapipe-model
This saves hand_landmarker.task to the current working directory. The demo nodes expect to find the file there at runtime.

​
AprilTag pick targets

The pick-and-place demo uses AprilTag pose estimates as pick targets. tag_1 is the default tag frame looked up by VacuumControl.get_tag_pose(). Print or fabricate a tag_1 marker from the 36h11 family and place it on the object or surface you want the arm to pick from.

​
Launch on real hardware

Once the arm, gripper, and camera are configured: 
pixi run launch
This is an alias for pixi run launch-lan, which builds the workspace if needed and then runs: 
ros2 launch ur_robot_driver ur_control.launch.py \
  ur_type:=ur3e \
  robot_ip:=192.168.56.101 \
  use_mock_hardware:=false \
  headless_mode:=true \
  launch_dashboard_client:=false \
  gripper_type:=vacuum

Build docs developers (and LLMs) love

Get started for freeTalk to us