Note: this post has been moved from my old website. Since then, the course has been archived, but I kept working until Project 4 so I may post videos of it later.

This is my homework for UC Berkeley CS188 (course archives), a course on Artificial Intelligence delivered by the University of California, Berkeley, and provided online by the EDX. I will complete this page as I continue following the courses’ archives.

For each project, a framework in Python was provided by the professor as a context to implement some AI. Students must implement key methods to fulfill specific requirements for the project (e.g. find the solution in fewer than 1000 iterations), and can fulfill optional objectives for a better grade (e.g. use fewer than 300 iterations). I don’t provide the source code of my implementation since it is supposed to be homework, but if you are personally interested in a specific algorithm, you can ask me for it (although there is nothing fancy in my answers, they do the job with a little optimization).

Project 1: Pacman search

In this project, we work with a Pacman clone where Pacman must eat all the pellets (there are no ghosts). We code an AI agent to control Pacman and solve the search problem of collecting all pellets following a certain heuristic.

One-Goal Problem – Big Level A* Search

Finding one pellet in a big maze

Corners Problem – Medium Level A* Search

Finding 4 pellets, each in one corner of a medium maze

Food Problem (Multi-Pellet Problem) – Tiny Level A* Search

Finding many pellets in a small level

Food Problem – Tricky Level A* Search

Finding many pellets in a level where they are placed in a tricky way

Project 2: Multi-Agent Search

Done, I will post videos later

Project 3: Reinforcement Learning

Done, I will post videos later

Project 4: Ghostbusters

Almost completed, I will post videos later