The Pac-Man Projects, developed at UC Berkeley, explore AI concepts by designing intelligent agents for the classic Pac-Man game. It serves as a hands-on platform for experimenting with search algorithms, adversarial strategies (like minimax), reinforcement learning, and probabilistic inference. These techniques allow Pac-Man to efficiently navigate mazes, evade ghosts, and maximize rewards while applying fundamental AI algorithms to real-world problems.
- Implemented depth-first, breadth-first, uniform cost, and A* search algorithms. These algorithms are used to solve navigation and traveling salesman problems in the Pacman world.
- Classic Pacman is modeled as both an adversarial and a stochastic search problem. Implemented multiagent minimax and expectimax algorithms, as well as designing evaluation functions.
- Implemented Reinforcement Learning algorithms to train agents in grid-based environments. The focus is on value iteration, Q-learning, and policy optimization techniques, applying them to solve tasks such as navigating mazes and collecting rewards.
- Implemented probabilistic tracking algorithms to enable an agent to locate hidden objects within a grid-based environment. It focuses on techniques like Hidden Markov Models and particle filters to estimate the positions of moving targets based on uncertain and noisy sensor data.