Project information

• Category: Artificial Intelligence
• Project Tools: Python
• Project date: November 2021
• Project URL: GitHub Repository

Project description

The Pac-Man projects were developed for UC Berkeley's introductory artificial intelligence course. These projects cover three main topics in artificial intelligence: search, multiagents, and reinforcement learning.

Search: Pac-Man lives in a shiny blue world of winding corridors and delicious food. Optimum movement in this world is Pac-Man's first step to success in this world. There are different scenarios in this project to aim the goal of optimizing the Pac-Man's movement: Find the fixed point using BFS, Implementation of DFS, Changing the cost function, A star search, Finding all the corners.

Multiagents: In this project, we design an agent for the classic Pa-Mman game, which, this time, also includes ghosts. In this path, we use minimax search and expectimax and design the evaluation function. Some scenarios are: Reflect agent, Minimax, Alpha-Beta pruning, Expectimax, Evaluation function.

Reinforcement learning: In this project, we implement value iteration and Q-learning. First, we evaluate our agents in Gridworld and then, apply that on a simulated robot, called Crawler, and the Pac-Man. Here, we can run the gridworld.py in mannual state and use the arrow keys to move on the space. Remember that if you press the up key, the agents obey it in 80% of the time. Scenarios in this part are: Value iteration, Bridge crossing analysis, Policies, Q-learning, Greedy epsilon.