🎮 AI Game Controller

Neuroevolved Agent for a 2D Platformer (C++ / SFML)

AI-Controlled Mode	Manual Mode

---

🧠 Overview

The AI Game Controller project is a refactored version of my Master’s dissertation that explores how an AI agent can learn to play a 2D platformer level in a way that resembles human gameplay.

The system uses a neuroevolutionary algorithm — evolving artificial neural networks through genetic selection — to teach an AI controller how to complete a Super Mario–style level built with SFML in C++.
The final challenge asks: Can observers tell if the player is human or AI?

---

🚀 Features

• 🧩 AI Controller: Learns movement and jump patterns using a genetic algorithm (NEAT-inspired).
• 🎮 Human Play Mode: Manual controls for testing and comparison.
• 🧱 Custom 2D Engine: Core loop, input mapping, physics, and collision detection built from scratch.
• 📊 Performance Logging: Option to output generation and fitness data for analysis.
• 🧾 Archived Build: Ready-to-play executable available for download.

---

📦 Download & Run

You can download the latest standalone executabke here:

👉 Download Standalone Executable

After downloading:

• Extract the .zip file.
• Open the folder and run the executable:
  • Windows: AI_Game_Controller.exe
• Use the instructions below to play through debug mode or toggle AI simulation mode.

---

🕹️ How to Start / Mode Select

• Launch the game.
• When prompted, use Up / Down to choose:
  • Play Manually (Debug Mode)
  • Start Simulation (AI Mode)
• Press Enter to confirm your selection.

---

🌍 Global Actions

Action	Key
Open / Close Pause Menu	Space
Exit Game (any time)	Esc

---

⏸️ Pause Menu

A global Pause Menu is available in both modes and functions just like the main menu.
Use Up / Down to navigate and Enter to select.

Options:

• Resume – return to the game.
• Quit – close the game entirely.

---

🎮 Manual Mode Controls

Action	Key
Move Left	←
Move Right	→
Look Up	↑
Crouch	↓
Run (hold to keep speed)	S
Jump	A
Spin Jump (SJUMP)	D

---

🤖 AI Mode

• In AI Mode, the controller automatically plays the game using neural-network-based behavior.
• Player input is disabled, but you can still global actions as required.

---

🧩 How It Works

1. The program creates a population of neural networks—each representing an AI controller.
2. Every controller plays the level and is scored by fitness (distance, survival time, completion).
3. The best controllers are selected, mutated, and crossed over to form the next generation.
4. Over time, the AI learns timing, jumping, and obstacle avoidance.
5. The evolved agent is tested against human gameplay for comparison.

---

🧱 Folder Structure

AI-Game-Controller\
|
SFMLGameEngine
+-- Bin/          # Downloadable builds (executables)
+-- Code/         # Source code (C++)
+-- Resources/    # Fonts, Sprites, Shaders, Outputs, etc.
+-- Previews/     # Screenshots, GIFs
\-- README.md     # Project documentation

---

🧠 Learning Objectives

• Design and implement a 2D platformer using SFML and modern C++ conventions.
• Apply neuroevolution (genetic algorithm + ANN) to gameplay control problems.
• Evaluate and compare AI and human playstyles using performance metrics.
• Refactor legacy academic code into a maintainable, modular structure.
• Log, visualize, and interpret AI learning results.

---

🧩 System Requirements

Platform	Minimum Spec
Windows	Windows 10+, SFML runtime included
CPU	Dual-core 2.0GHz
RAM	4GB
GPU	Any with OpenGL 2.1 support

---

🧱 Game States

State	Description
TitleState	Displays “Press Any Key To Start” and transitions to the Main Menu upon key press.
MainMenuState	Title and mode selection (Manual or AI).
LoadingState	Preloads textures, world objects, and assets before the game begins.
MainState	Active gameplay, player and AI logic, and camera management.
PauseMenuState	Opens with Space, offering Resume, Title, or Quit options.

---

🎬 Splash Screens

Title Screen	Main Menu
Manual Gameplay	Auto Gameplay
Pause Menu

---

📚 Future Enhancements

• Add multiple level files and adjustable parameters.
• Implement save/load for trained neural networks.
• Display visual AI debugging overlays.
• Package cross-platform release builds.

---

👤 Author

Mohamed Agilah
🎓 Games Programmer & AI Developer
🌐 Portfolio Website
📧 Contact: agilahmohamed@gmail.com

Project archived for educational and portfolio purposes (October 2025).