- Please visit our website for more information about LEAP hand.
- Please see the Python API, ROS API, ROS2 API, Useful Tools folders for software specific details.
- Connect 5v power to the hand (the dynamixels should light up during boot up.)
- Connect the Micro USB cable to the hand (Do not use too many USB extensions)
- Open Dynamixel Wizard and find the correct port using the options button and put that in main.py or ros_example.py. Note, you cannot have Dynamixel wizard open while using the hand's API, the port will be "busy" with the other process.
- On Ubuntu you can find the hand by ID using
/dev/serial/by-id
The ID will stay persistent on reboots. - We offically support Python and ROS, but other languages are supported by Dynamixel SDK.
- To improve latency on Ubuntu try these tips. Configure USB Latency Settings in Ubuntu and the Dynamixel Python SDK
- Leap Node allows you to command joint angles in different scalings.
- You can read position, velocity and current from the hand.
- Do not query reads too often, going past 90hz for one set of angles or 30hz for all three will slow down the USB communication.
- The default controller follows the PID control, up to the current limit cap.
- Other controllers including velocity control or current control are supported as per the motor manual
- For Lite, keep the current limit around 300ma.
- For Full, you can raise the current limit up to 550ma.
- If facing a jittery hand, adjust the PID values down.
- If the hand is too weak, adjust the PID values up.
- If your motor is 90/180/270 Degrees off, the horn is mounted incorrectly on the motor. Remount it.
- If no motors show up, check that your serial port permissions are correct.
- If some motors are missing, make sure they are IDed corrrectly and are connected to the U2D2.
- If you get "overload error" and the motors are flashing red, then they have overloaded (self-collision etc). It should clear on a power cycle. If it happens often, lower the current limits in the control code so that it does not happen as often.
- If you get "jittery" motors, try lowering the P and D values, either in the roslaunch file or the python file.
- If you feel the motors are too inaccurate, you can also try raising the P and D values.
- MANO to LEAP joint angle mapping.
- If you have useful tools you feel the community can benefit from, please make a pull request.
- Please see the code from Bimanual Dexterity for Complex Tasks to use Manus gloves with LEAP Hand.
- I can also add tools to this upon request. :)
- Please contact me at kshaw2@andrew.cmu.edu for any issues.
- This code is made available using an MIT License.
- The CAD files are provided with a CC BY-NC-SA Attribution-NonCommercial-ShareAlike license which allows you to use and build upon our work non-commercially.
- LEAP Hand is provided as-is and without warranty.
- If you use LEAP Hand in an academic setting, please cite our paper:
@article{shaw2023leaphand,
title={LEAP Hand: Low-Cost, Efficient, and Anthropomorphic Hand for Robot Learning},
author={Shaw, Kenneth and Agarwal, Ananye and Pathak, Deepak},
journal={Robotics: Science and Systems (RSS)},
year={2023}
}