UR-5 Ball in Cup Environment

This repository contains the code to use RL on a UR5 robot with the traditional ball-in-cup game. The necessary package and basic instructions are presented. Before running this environment, ensure that the following requirements are met:

1. Clone and Install Python URX Package

1. Clone the Python URX repository from GitHub:

git clone https://github.com/UoA-CARES/python-urx.git

2. Navigate into the cloned repository directory:

cd python-urx

3. Install the Python URX package in editable mode using pip:

pip install -e .

2. Install Math3d Package

1. Install the math3d package version 3.4.1:

pip install math3d==3.4.1

3. Install Pypylon

1. Install this library, need for Basler Camera

pip install pypylon

Connect the camera to a USB 3.0 port.

Test the simulation

A helpful Docker file is available for testing ideas in simulation. Simulation provides a safe environment to experiment with new ideas without risking the real robot. So highly recommended to uses. In a new terminal run:

sudo docker run --rm -it   -p 6080:6080 -e ROBOT_MODEL=UR5 -v "${HOME}/ur_programs:/ursim/programs" universalrobots/ursim_e-series

Then open a web browser and past the following link and press connect

http://172.17.0.2:6080/vnc.html?host=172.17.0.2&port=6080

You will be able to view the control screen panel exactly as it appears on the real robot. Simply follow the same initialization setup used for the real robot. Make sure to use the right IP address in the robot_config.json. This simulation will not replicate the entire RL environment. Specifically, the ball (or cube) will not be present in the simulation. It focuses solely on the robot itself, without any external environmental influences.

Run the provided train.py example to ensure everything is set up correctly:

python train.py

Test the real robot

Turn on the real robot and follow the initialization instructions displayed on the control panel screen. Ensure that the robot is in the safe operational area before running any code. Make sure to use the right IP address in the robot_config.json

Be careful here; the robot will start to move

Run the provided train.py example to ensure everything is set up correctly:

python train.py

Also, I have hard-coded the task solution, which consists of a sequence of movements to place the cube inside the cup. Simply call the function "hard_code_solution" located in the "main_environment_ur5.py" file.

To-Do List

Here are some tasks that need to be completed to finish or improve this project:

• Update the code to ensure compatibility with the CARES RL package. You can find the package here
• Improve the camera setup. Although the camera calibration has been done, it is highly recommended to recalibrate the camera in the robot's area for better accuracy.
• Enhance the visualization of the task execution. Currently, only the cube and the ArUco marker are displayed. It should display additional information similar to what is done for the 2-finger gripper. You can find an example here