mimic_hand_middleware 0.3.0.0

Joint-to-motor level control API for mimic robotics AG's robotic hands.

mimic_hand_middleware

Repository for the joint-to-motor level control of mimic's robotic hands.

System specifications

We support the following platforms:

• x86 architecture
• Ubuntu Jammy (22.04)
• 2 USB-A ports

Install

To install the hand API, clone the this repo and run its install script:

mkdir ~/git && cd ~/git
git clone [e-mail]:mimicrobotics/mimic_hand_middleware.git
./scripts/install.sh

Since the script installs new udev rules for the hand, when prompted, you'll need to enter a sudo password. The new python environment will be installed in $HOME/python_venvs/mimic_hand, so when opening a new terminal, make sure to activate this environment by running:

source ~/python_venvs/mimic_hand/bin/activate

When connecting the hand's USB board to your PC, Ubuntu will set the latency timer of the async buffer to 16 ms. This is too slow for our purposes, so we override it by running:

./scripts/connect_mimic_hand.sh

Optionally, you can paste the lines above into your ~/.bashrc file - this will source your environment and modify the latency timers every time you open a new terminal.

Run

First, you can run a test sequence on all of the joints by running:

source ~/python_venvs/mimic_hand/bin/activate
python tests/test_joint_cmd.py

You can also move the hand around yourself in a simple GUI:

source ~/python_venvs/mimic_hand/bin/activate
python python tests/grasp_gui.py

Using our API

The main interface to our hands is the GripperController class. You can initialize it as follows:

from mimic_hand_middleware import GripperController
gc = GripperController()
gc.connect_motors() # connect, set maximum current and operation mode (current-limited position control)
gc.init_joints(calibrate=False) # enable torques on the motors

Sending joint commands to the hands

Motor commands are sent to all motors at once, as a numpy array of shape (16,). Commands are sent in degrees. To have a look at the zero positions, you can start up the grasp_gui, and select joint control mode to home all joints to 0.

# commanding joint angles in degrees
gc.command_motor_angles(motor_cmd_array)

The numbering of joints follows a simple convention: idx 0 starts from the base of the thumb, then indexes increase with the distance from the finger base until 3 on the thumb, then on the index etc.

Reading motor angles, speeds and currents from the hands

You can read the current motor angles, speeds and currents as follows:

motor_status = self.gc.get_motor_pos_vel_cur()

Where motor_status is a list of length of 3, containing numpy arrays for motor positions (rad) speeds (rad/s) and currents (mA).