Robotic In-Hand Manipulation for Large-Range Precise Object Movement: The RGMC Champion Solution

[Project Website]

Repository for the paper Robotic In-Hand Manipulation for Large-Range Precise Object Movement: The RGMC Champion Solution, IEEE Robotics and Automation Letters, 2024.

In this repository, we provide:

• The algorithm for our proposed in-grasp object movement approach.
• A simulator based on MuJoCo for a quick test of the algorithm.

We do not provide code related to the hardware implementation, as it depends on your specific hardware setup.

Installation

1. Install ROS Noetic.

2. Create a conda virtual env and install Pinocchio:

   conda create -n <YOUR_ENV_NAME> -c conda-forge python=3.9 pinocchio

3. In your python env:

   pip install mujoco==3.1.3
   pip install dm_control==1.0.16
   pip install rospkg

4. Clone this repo into a ROS workspace.

5. Build the workspace:

   cd <YOUR_WORKSPACE>
   catkin build
   

Usage (Simulation)

Run the mujoco simulation only (no motion):

# in your python env
cd leap_task_A/scripts
source ../../../../devel/setup.bash
python leaphand_mujoco.py

Run the in-grasp object movement:

# in your python env
cd leap_task_A/scripts
source ../../../../devel/setup.bash
python leaphand_control.py