Neuro Gaming Lab

Neuro-Adaptive-Ghosts-ML

Where nostalgia meets intelligence

Reimagining Pacman with adaptive ghost AI—coordinated pursuit, alpha pack hunting, RL, and evolution. Pygame + ML.

ML RL Ghost AI Coordinated agents

We reimagine Pacman as a living maze—where ghosts use ML to coordinate, pick an alpha leader, block escape routes, and evolve over time. Classic chase, reimagined with reinforcement learning, multi-agent coordination, and genetic evolution.

Adaptive ghost agents: role-based strategies (chaser, ambusher, blocker, patrol), alpha pack hunting when you’re close, optional PPO/DQN/A2C policies, and evolutionary behavior tuning. Pacman as a testbed for emergent coordination and playable AI.

From Pygame gameplay to trained models and in-game evolution, Neuro Gaming Lab explores how learning agents master and meaningfully interact with games. Reviving the past to build the future of playable AI.

Features

2D gameplay

Pacman ML gameplay 1 — coordinated chase
OBSERVE: Agentic Ghosts — coordinated chase
Pacman ML gameplay 2 — alpha pack
Alpha pack hunting & roles
Pacman ML gameplay 3 — power pellet
Power pellet & ghost evolution
Pacman ML gameplay 4 — maze
Neural pathfinding & maze

Listen

Neuro Adaptive Ghosts ML — Arcade Classics

3D sphere view

Run with --sphere-view to play on a 3D sphere: the maze wraps onto the surface and the camera follows Pacman. Hunting lines show ghost targets in 3D.

Pacman 3D sphere 1 — play surface
3D sphere — play surface
Pacman 3D sphere 2 — camera follows
Camera follows player
Pacman 3D sphere 3 — hunting lines
Hunting lines on sphere
Pacman 3D sphere 4 — wrap and chase
Wrap & chase in 3D

Listen — 3D Spherical

Surviving Adaptive Ghosts ML in Spherical — Arcade Classics

Run the game

Clone the repo, install dependencies, and run the full game (Pygame) or the web version. Use --sphere-view for the 3D sphere playground (requires PyOpenGL).