zephyr0402 / reinforcement-learning-with-nn

Trains a robot in a robocode using Q-learning and Neural Networks

This code trains a robot in a robocode environment using Reinforcement Learning approach in specific Q-learning algorithm. (For more information about robocode see Robocode.

Overview

LookUpTable

NeuralNetwork

Note: The neural network learns Q-value and the corresponding state action pair online.

Getting Started

1. Install Robocode (check link for installation )
2. Make sure you include your robocode environment variable in your IDE
3. Put the robots either in the robot folder (The robot folder can be found under the directory where robocode is installed)
4. Start the battle with the one of the RL robot and choose an enemy robot.
5. While testing set the greedy boolean variable to true

Dependencies

• Robocode
• Java
• Eclipse (or other IDE, make sure you have added Robocode environment variable) Note: Change the package name to the one you have created.