Usage Instructions

Installing Python Driver for T-Motor AK80-6 Actuator

  1. Clone the python motor driver from: https://github.com/dfki-ric-underactuated-lab/mini-cheetah-tmotor-python-can

  2. Modify the .bashrc file to add the driver to your python path. Make sure you restart your terminal after this step.:

    # mini-cheetah driver
export PYTHONPATH=~/path/from/home/to/underactuated-robotics/python-motor-driver:${PYTHONPATH}

Make sure you setup your can interface first. The easiest way to do this is to run sh setup_caninterface.sh from the mini-cheetah-motor/python-motor-driver folder. To run an offline computed swingup trajectory, use: python3 swingup_control.py. The script assumes can id as ‘can0’ and motor id as 0x01. If these parameters differ, please modify them within the script. Alternatively, the motor driver can also be installed via pip from https://pypi.org/project/mini-cheetah-motor-driver-socketcan/.:

pip install mini-cheetah-motor-driver-socketcan

Setting up the CAN interface

  1. Run this command and make sure that can0 (or any other can interface depending on the system)shows up as an interface after connecting the USB cable to your laptop: ip link show

  2. Configure the can0 interface to have a 1 Mbaud communication frequency: sudo ip link set can0 type can bitrate 1000000

  3. To bring up the can0 interface, run: sudo ip link set up can0

    Note

    Alternatively, one could run the shell script setup_caninterface.sh which will do the job for you.

  4. To change motor parameters such as CAN ID or to calibrate the encoder, a serial connection is used. The serial terminal GUI used on linux for this purpose is cutecom

Testing Communication

To enable one motor at 0x01, set zero position and disable the motor, run: python3 can_motorlib_test.py can0

Use in Scripts: Add the following import to your python script: from canmotorlib import CanMotorController after making sure this folder is available in the import path/PYTHONPATH.

Example Motor Initialization:

`motor = CanMotorController(can_socket='can0', motor_id=0x01, socket_timeout=0.5)`

Available Functions:

  • enable_motor()

  • disable_motor()

  • set_zero_position()

  • send_deg_command(position_in_degrees, velocity_in_degrees, Kp, Kd, tau_ff):

  • send_rad_command(position_in_radians, velocity_in_radians, Kp, Kd, tau_ff):

All functions return current position, velocity, torque in SI units except for send_deg_command.

Performance Profiler: Sends and received 1000 zero commands to measure the communication frequency with 1/2 motors. Be careful as the motor torque will be set to zero.

Using different Controllers for the Swing-Up

All implemented controllers can be called from the main.py file. The desired controller is selected via a required flag, e.g. if you want to execute the gravity compensation experiment the corresponding command would then be:

python main.py -gravity

Make sure you execute the command from the directory /software/python/examples_real_system otherwise you have to specify the path to main.py as well. If you want to autosave all data from your experiment use the optional flag -save together with the flag required for the controller.

To get an overview of all possible arguments display help documentation via:

python main.py -h

Saving Results

The results folder serves as the directory, where all results generated from the python code shall be stored. The distinct seperation between python script files and generated output files helps to keep the python package clear and tidy. We provide some example output data from the very start, so that you may see what results each script produces even before you run the code. The tools to create result files from the respective experiment data are located within the python package under /utilities, plot.py, plot_policy.py and process_data.py.

In general particular functions get called from main.py or another script to produce the desired output, which improves resuability of the utilities and keeps the code concise. The results of each experiment are saved in a new folder, which is automatically assigned a timestamp and an appropriate name.