AI Bot Development Hackathon

Colab Notebook

You can view the example notebook here:
Colab Notebook

Project Overview

This hackathon focuses on developing AI bots that can play a 2D game using reinforcement learning. Participants will create their own bot implementations and reward functions to compete in a dynamic game environment.

Environment Setup

1. Clone the repository
2. Install dependencies:

   pip install -r requirements.txt

Project Structure

├── bots/                    # Bot implementations
│   ├── simple_example_bot.py # Example bot implementation
│   └── your_bot.py         # Your bot implementation
├── components/             # Game components
├── environment.py          # Game environment
└── main.py                # Main training loop

Hackathon Challenge

Overview

Your task is to develop an AI bot that can play a 2D game by implementing:

1. The reward function in the environment
2. Your own bot implementation

Environment Features

The game environment (environment.py) provides:

• A 2D game world with obstacles
• Player characters that can move, rotate, and shoot
• State information about the game
• A reward calculation system

Available Information

Your bot receives the following state information:

• Player position (x, y)
• Player health
• Player angle
• Player velocity (x, y)
• Closest opponent position
• Other game state information

Available Actions

Your bot can perform these actions:

• Move forward/backward/left/right
• Rotate
• Shoot

Implementation Guide

1. Reward Function

Implement the reward function in environment.py to define good and bad behavior. Consider rewarding:

• Staying alive
• Dealing damage to opponents
• Getting kills
• Moving efficiently
• Avoiding obstacles

2. Bot Implementation

Create a new bot class in the bots directory. Your implementation can include:

• Neural networks
• Experience replay
• Action selection logic
• State processing
• Or any other approach you choose (there are no restrictions on the architecture), it can even not be a neural network

Example Implementation

A simple example bot (simple_example_bot.py) demonstrates:

• Basic neural network structure
• Simple state processing
• Random action exploration
• Experience replay implementation

Getting Started

1. Create a new file in the bots directory for your bot
2. Implement your bot class with the required methods:

   class YourBot:
       def __init__(self, action_size):
           # Initialize your bot
           pass
   
       def act(self, state):
           # Return actions based on state
           pass

3. Modify the reward function in environment.py

Tips for Success

• Start with a simple reward function and gradually make it more sophisticated
• Use the example bot as a reference but try to improve upon it
• Consider adding more features to your neural network
• Experiment with different hyperparameters
• Test your bot's performance against the example bot

Good luck with the hackathon!