This repository contains a C++ implementation of the Q-Learning algorithm applied to a simple grid world problem. The grid world consists of states represented by rows and actions represented by columns. The goal is to reach a specific state denoted by F from any staring point.

Here is the representation of the environment:

Here is the Reward Table:

The q_learning.cpp file contains the main implementation of the Q-Learning algorithm. Here's an overview of the key components:

• printArray: Function to print a 2D array.
• initializeRewardArray: Function to initialize the reward matrix based on the problem description.
• getRandomPossibleAction: Function to select a random possible action from a given state.
• QLearningAlgorithm: Function implementing the Q-Learning algorithm. It updates the Q-table based on the rewards and learns the optimal policy.
• main: The entry point of the program. It initializes the reward matrix, sets up the Q-table, runs the Q-Learning algorithm, and prints the final Q-table.
  
  

You can modify the parameters such as the number of episodes (episode) and the learning rate (y) in the main function to experiment with different settings.

// The # of trials for computer to learn, usually the # is big
int episode = 1000;

// Constant for learning rate
double y = 0.8;

Here is the Q matrix after 1000 episodes: