Robotic-Arm-Control-in-Table-Tennis-with-Reinforcement-Learning

This project aims to develop and train a neural networks to control a robotic arm for playing table tennis in a simulated environment in PyBullet. The project employs supervised learning for inverse kinematics and reinforcement learning to optimize the robot’s paddle movements.

How to Run:

To run the server: python .\src\server.py
All the possible options for the server are avaiable at interface.txt.

To play with the trained model:
python .\src\test.py \

Supervised Learning:

To start a Supervised learning session:

• High Player: python .\src\train_supervised_high.py;
• Low Player: python .\src\train_supervised_low.py;

Reinforcement Learning:

To start a reinforcement learning session, you need to start the server and connect two players:

• 1 player: python .\src\server.py -auto;
• Player to train: python .\src\train_reinforcement.py;