Maze Solver with Deep Q-Network (DQN)

This project implements a Deep Q-Network (DQN) agent to solve a maze environment. The agent is trained to navigate through the maze to reach a goal position, using reinforcement learning techniques.

Project Structure

• maze.py: Defines the Maze class, representing the maze environment.
• DQNAgent.py: Defines the DQNAgent class, implementing the DQN algorithm.
• dqn_main.py: Main script to train the DQN agent and evaluate its performance.
• dqn2.ipynb: Jupyter notebook for training and visualizing the DQN agent's performance.
• rl_evaluator.py: Custom module for evaluating reinforcement learning agents.

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/maze-dqn.git
   cd maze-dqn

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install the required packages:

   pip install -r requirements.txt

Usage

Training the DQN Agent

To train the DQN agent, run the dqn_main.py script:

python [dqn_main.py]

This script will train the agent over a specified number of episodes and plot the rewards and losses after training.

Evaluating the Agent The evaluate_agent function in dqn_main.py can be used to evaluate the agent's performance over a specified number of runs. The function returns the average reward over the evaluation runs.

Visualizing the Agent's Path The plot_path function in dqn_main.py can be used to visualize the path taken by the agent in the maze. The function plots the start, end, and a few intermediate points along the agent's path.

Jupyter Notebook The dqn2.ipynb notebook provides an interactive environment for training and visualizing the DQN agent's performance. You can run the notebook to see detailed plots and animations of the agent's learning process.

Maze Environment The maze environment is defined in maze.py. The Maze class provides methods for creating the maze, resetting the state, taking steps, and computing rewards.

Maze Level

The maze level is defined as follows:

level = [
    "XXXXXXXXXXXXX",
    "X           X",
    "X XXX X XXX X",
    "X   X X   X X",
    "XXX X XXX X X",
    "X   X   X   X",
    "X XXX XXX X X",
    "X X   X   X X",
    "X XXX X XXX X",
    "X   X X   X X",
    "XXX XXX XXX X",
    "XP        X X",
    "XXXXXXXXXXXXX",
]