Skip to content

Latest commit

 

History

History
38 lines (24 loc) · 1.81 KB

README.md

File metadata and controls

38 lines (24 loc) · 1.81 KB

Grasp-with-the-Unitree-D1

Demonstrations of Using the Unitree D1 Robotic Arm to Open Cabinet Doors and Grasp Objects

Click here to watch the video

Grasping Scissors (5x Speed)
demo-1

Grasping Phillips Screwdriver (Tool selected based on real-time audio recognition)
demo-2

Grasping Flathead Screwdriver
demo-3

Opening Drawer
demo-4

Usage Example: src/test_grasp.py

How to Build from Scratch?

  1. Use the script src/get_arm_joint_angle.cpp to record key poses for grasping:

    1. After running it in the terminal, you can get real-time joint angles. On MAC, use the shortcut by pressing 'r' followed by 'ENTER' to record.
    2. Key poses will be saved in build/servo_commands.txt, which can be copied into multi-joint angle control programs, such as left_1.cpp.

  2. Implement multi-joint angle control for grasping, split into multiple stages (e.g., opening the left cabinet door is split into left_1.cpp and left_2.cpp) because:

    1. It makes debugging easier;
    2. Increases the accuracy of grasping;
    3. Helps alleviate communication message delay (alternatively, add a sleep command or wait for the previous message to execute before sending the next one).

  3. Ensure that the angle of the sixth joint remains unchanged after grasping the tool until it is released, to prevent it from dropping.

  4. Modify CMakeLists.txt, use cmake to build, and make to compile into binary shared libraries, such as libleft_1.so and libleft_2.so.

  5. Call the shared libraries in C++ or Python to implement grasping. Example usage can be found in src/test_grasp.py.