Revo 2 Adaptation for Unitree G1 (Testing & FAQ)

Verification & Testing

To perform verification testing, launch two separate terminal instances simultaneously:

Terminal 1: Launch Controller Node and Stark Node

conda activate g1brainco                    # Activate the conda environment
cd ~/unitree-g1-brainco-hand/brainco_ws     # Navigate to the workspace
./launch/launch_robot.sh                    # Execute the launch script

Verify the terminal outputs:

• Confirm that the Port, Baudrate, and slave_id are correct for both hands.

• Check for "serial port opened" to ensure successful serial communication.

• Check for "Waiting for joint cmd ..." indicating the node is awaiting command inputs. If the above prints do not appear, verify the configurations in your parameters file.

• Check for "IK initialization done." indicating inverse kinematics setup was successful.

• When "Request 'configure' to start" is displayed, the system is ready to accept transition requests.

Terminal 2: Launch State Machine Transition Node

conda activate g1brainco                    # Activate the conda environment
cd ~/unitree-g1-brainco-hand/brainco_ws     # Navigate to the workspace
./launch/launch_trans.sh                    # Execute the launch script

Triggering State Transitions

Terminal 2 will display the current active state, list available transitions, and show their corresponding actions. Type the designated character(s) + Enter to trigger state transitions. Once the system enters the active state, you can append l or r after command characters to control only the left or right hand respectively. Omitting the suffix defaults to controlling both hands simultaneously.

Hand-Eye Calibration

This procedure supports G1-29 DOF only. Upper-body motions do not lock WaistRoll and WaistPitch; use single-waist mode or add the required locking logic before calibration.

1. Robot Configuration

Modify the general configuration in brainco_ws/src/control_py/config/smach_config.yaml:

mode:
  current_mode: calibr # Select hand-eye calibration mode

Check the calibration configuration in brainco_ws/src/control_py/control_py/state_manager/calibrate/calibrate_config.yaml:

paths:
  cam_calibr_dir: /home/unitree/unitree-g1-brainco-hand/cam_calibr

collection:
  move_range: [0.015, 0.015, 0.015, 0.15, 0.15, 0.15]

• cam_calibr_dir: Directory for saving calibration data
• move_range: Random end-effector motion range during calibration data collection

2. Unitree Terminal 1: Start the Main Control Node and Dexterous Hand Node

# Enter the workspace
cd ~/unitree-g1-brainco-hand/brainco_ws
# Activate the environment
conda activate g1brainco
# (Recommended) Start arm-only control
./launch/launch_robot.sh arm

Check the output:

• IK initialization is complete: "IK initialization done."
• When "Request 'configure' to start" is displayed, you can send state-transition requests.

3. Unitree Terminal 2: Start the Agent Network Service

# Enter the workspace
cd ~/unitree-g1-brainco-hand/brainco_ws
# Activate the conda environment
conda activate g1brainco
# Start the Agent service
./launch/launch_trans.sh agent

4. Client PC: Open the Control GUI

# Enter the client directory
cd ui_client
# Activate the conda environment
conda activate braincogui
# Start the GUI
python ui_client/ui_client.py

5. Collect Hand-Eye Calibration Data

Prepare a chessboard calibration board first.

• Calibration board bracket: Download and print the calibration board bracket, which uses the same wrist interface as Revo 2.
• Chessboard calibration board: Produce the board from the calibration board pattern, then mount it on the bracket.

After assembly, you can directly replace the Revo 2 hands with the board/bracket assembly for calibration.

Use the following state sequence to collect data for the left and right arms:

Init -> Ready -> Active -> Calibr-Left-Start -> Calibr-Right-Start -> Ready

During collection, the program automatically moves the robot arm randomly and saves the data. By default, 20 samples are collected for each side.

Generated data files:

• Images: cam_calibr/imgs/left/, cam_calibr/imgs/right/
• Robot arm poses: cam_calibr/arm_pose/left.json, cam_calibr/arm_pose/right.json

6. Compute Calibration Results Offline

After collecting data for both left and right arms, check the images:

cd ~/unitree-g1-brainco-hand/cam_calibr
chmod +x ./imgs/check_img_grid.sh
./imgs/check_img_grid.sh left
./imgs/check_img_grid.sh right

Compute the calibration parameters offline:

cd ~/unitree-g1-brainco-hand/cam_calibr
python compute_calibration.py

The script writes the calibration result to cam_calibr/settings.yaml.

This file contains the camera extrinsics for the left and right hands. The calibrate mode reads this file later for coordinate conversion.

---

FAQ

1. Custom Network Configuration Restricts Internet Access

If configuring a custom static IP breaks internet connectivity, delete the default interface gateway routing rule:

sudo ip route del default via 192.168.123.1 dev eth0

2. Colcon Build Displays System Clock Skew Warnings

make[2]: warning:  Clock skew detected.  Your build may be incomplete.

This warning occurs when the system time defaults to 1970 due to a lack of network connection for automatic NTP synchronization. Solution:

1. Remove the default routing rule restricting internet access:

sudo ip route del default via 192.168.123.1 dev eth0

2. Restart the terminal and verify the system time and synchronization status:

date
timedatectl status

3. Once the time is successfully synchronized, clean up previous build caches and rebuild the workspace:

rm -rf build install log
colcon build

3. "ros2: command not found" or "Package 'XXX' not found"

This indicates that the ROS 2 overlay or underlay environment configurations have not been sourced in the current shell. Make sure to run:

source /opt/ros/foxy/setup.bash
# or source your active workspace's setup script
source install/setup.bash

4. How to Resolve General ROS 2 Execution Failures

If you encounter persistent ROS 2 node crashes or runtime exceptions:

1. Close all active terminal sessions and open clean terminals.
2. Clean up previous build outputs (rm -rf build install log) and rebuild the workspace.

5. Wi-Fi Card Interface Blocked (RF-Kill Issue)

unitree@ubuntu:~$ sudo ifconfig wlan0 up 
# Error: SIOCSIFFLAGS: Operation not possible due to RF-kill

Solution:

1. Check the hard/soft block status of wireless devices:

unitree@ubuntu:~$ rfkill list
# 0: phy0: Wireless LAN
#         Soft blocked: yes
#         Hard blocked: no

2. Unblock the Wi-Fi interface using rfkill:

sudo rfkill unblock wifi

3. Verify that the soft block status has changed to no and bring the interface up:

rfkill list
sudo ifconfig wlan0 up

6. VS Code Fails to Establish SSH Remote Connection

Ensure that the host device has internet connectivity. A typical failure is the remote extension host installer failing to download required VS Code plugins due to DNS resolution failures:

Failed to connect to the remote extension host server (Error: ExtensionInstallFailed(ExtensionInstallFailed("Error while installing extensions: getaddrinfo EAI_AGAIN marketplace.visualstudio.com...")

Solution:

1. Connect to the robot via standard shell SSH:

ssh unitree@192.168.xx.xx

2. Verify external domain connectivity:

ping marketplace.visualstudio.com

3. Check active DNS servers:

cat /etc/resolv.conf

If Ubuntu resolves DNS using the systemd-resolved local stub (127.0.0.53) which fails to forward upstream queries, configure static DNS servers. 4. Open the systemd-resolved configuration file:

sudo vim /etc/systemd/resolved.conf

Uncomment and edit the following properties:

[Resolve]
DNS=8.8.8.8 1.1.1.1 
FallbackDNS=114.114.114.114 
DNSStubListener=yes

5. Save changes and restart systemd-resolved:

sudo systemctl restart systemd-resolved

6. Verify active status and DNS mappings:

systemd-resolve --status

Ensure 8.8.8.8 and 1.1.1.1 are mapped as active servers. You can now establish VS Code SSH sessions successfully.

---

Contact & Support

For questions or purchase inquiries, contact the BrainCo Dexterous Hand Team or submit a support ticket through the page help entry.