Q-Learning bot Example (very simple)

[Deep Slayer 34]

Q-Learning OSRS Bot: Kills Cows & is punished for looting

Overview

This Old School RuneScape (OSRS) bot leverages Q-learning, a type of reinforcement learning, to fight cows. The bot focuses on optimizing two primary actions: attacking cows and avoiding the collection of cowhides. It learns and adapts over time to maximize rewards for attacking cows and minimizes penalties for picking up cowhides. 

Features

• Q-Learning Algorithm: Utilizes Q-learning to dynamically adapt to the game environment.
• State Management: Recognizes two primary states - Fighting and Looting.
• Reward System: Rewards attacking cows and penalizes picking up cowhides.
• Epsilon Decay: Ensures the bot transitions from exploring different actions to exploiting the most rewarding actions over time.
• Logging and Monitoring: Detailed logging to track actions, rewards, Q-value updates, and epsilon decay.

Technical Details

• Q-Table Storage: Saves and loads Q-values to persist learning across sessions.
• Adaptive Behavior: The bot learns from interactions, adjusting its behavior to optimize for long-term rewards.

How It Works

1. Initialization: The bot loads or initializes a Q-table to store Q-values representing the expected rewards of actions in various states.
2. State Recognition: Determines the current state (Fighting or Looting) based on the player’s status.
3. Action Selection: Chooses actions based on the current state, either attacking cows or looting cowhides, with a preference for actions with higher Q-values.
4. Reward System: Receives rewards for attacking cows and penalties for looting cowhides, adjusting Q-values accordingly.
5. Epsilon Decay: Gradually reduces exploration over time, focusing on the most rewarding actions.

How To Use

1. Start the bot near Cows

https://github.com/deepslayer/Q-LearningExample

State.java LearningBot.java LearningAgent.java Action.java

Edited June 11 by Deep Slayer