sabamadadi / wampus-game

Implementation of the Wumpus Game using reinforcement learning!

Wumpus World Environment Documentation

Introduction

Welcome to the Wumpus World Environment! This environment is designed to simulate a simple yet challenging grid-based game, where an agent navigates through a world containing dangers such as a Wumpus and pits while trying to find gold.

Environment Overview

The Wumpus World is a grid-based environment with the following key components:

• Agent: Represented by an icon, the agent explores the grid and makes decisions to avoid dangers and reach the gold.

• Wumpus: A dangerous creature that emits a stench. The agent must avoid encountering the Wumpus.

• Gold: The ultimate goal for the agent is to find and grab the gold.

• Pits: Perilous locations that emit breezes. Falling into a pit results in a penalty for the agent.

• Arrows: The agent has a limited number of arrows that can be used to shoot the Wumpus in a specific direction.

Actions

The agent can take the following actions:

• Move Up
• Move Down
• Move Right
• Move Left
• Shoot Up
• Shoot Down
• Shoot Right
• Shoot Left

The shooting actions consume arrows and eliminate the Wumpus if it is in the targeted direction.

State Representation

The state is represented as a tuple containing the following information:

• Agent's x-position
• Agent's y-position
• Breeze indicator (boolean)
• Stench indicator (boolean)
• Remaining arrows

Rewards

The agent receives rewards and penalties based on its actions and interactions with the environment:

• -10: Penalty for falling into a pit or encountering the Wumpus.
• +10: Reward for grabbing the gold.
• -1: Small penalty for each step taken.

Usage Instructions for Students

1. Environment Creation:

   env = WumpusWorldEnv(size=4, number_of_pits=3)

2. State Representation:

   • The state is a tuple containing the agent's position, breeze and stench indicators, and remaining arrows.

3. Actions:

   • The agent can take actions using integers from 0 to 7, each corresponding to a specific move or shooting direction.

4. Resetting the Environment:

   • Call env.reset() to start a new episode. This resets the agent's position, remaining arrows, and other parameters.

5. Taking Actions:

   • Use env.step(action) to perform an action in the environment. It returns the next state, reward, episode completion status, and additional information.

6. Rendering:

   • The environment supports rendering using Pygame. Call env.render() to visualize the current state of the grid.

7. Creating an Agent:

   • Students should implement an intelligent agent that decides the next action based on the current state.

8. Exploration and Decision Making:

   • The agent needs to explore the environment, consider the breeze and stench indicators, and decide on the best actions to avoid dangers and reach the gold.

9. Implement Intelligence Approach:

   • Implement an algorithm or strategy for the agent to make decisions. Consider using techniques like Q-learning, reinforcement learning, or other intelligent approaches.

10. Iterative Improvement:

    • Experiment with different intelligent approaches, tune parameters, and iterate to improve the agent's performance in the Wumpus World.