This repository contains multiple AI-powered Pacman projects developed using Python. The goal is to implement and explore fundamental AI concepts such as informed search, multi-agent decision-making, and probabilistic inference — all in the context of the classic Pacman game.
- Implemented A* search with custom heuristics.
- Designed agents to find optimal paths and efficiently collect food in a maze.
- Explored suboptimal strategies like greedy food collection.
- Built agents using Minimax, Alpha-Beta Pruning, and Expectimax.
- Developed evaluation functions for smart decision-making.
- Pacman plays against ghost agents that simulate adversaries or randomness.
- Designed probabilistic agents to track and locate invisible ghosts using noisy distance sensors.
- Implemented exact inference and belief updates over time.
- Explored real-world AI concepts like Bayesian reasoning and particle filters.
- Implemented model-free reinforcement learning agents (Q-Learning and SARSA) for Pacman, including ε-greedy exploration and temporal difference updates.
- Designed and tuned feature-based value function approximations to improve generalization across unseen game states.
- Evaluated agent performance under stochastic environments, analyzing convergence behavior, policy stability, and reward optimization.
- Python 3
- Grid-based game engine (provided by UC Berkeley Pacman Projects)
- Autograder scripts for evaluating correctness and efficiency
- Command-line based testing with optional graphical output
- Clone the repository:
git clone https://github.com/yourusername/artificial-intelligent-pacman.git cd artificial-intelligent-pacman