PhD Thesis Defense

I defended my PhD in 2011, may the 9th. You can find the slides in this section.

PhD Thesis Manuscript

You can find here the last version of my PhD thesis manuscript.

Research Papers

You can find here my research papers.

YASkySim - Skywar Simluator

I created this simulator for the last game by Motion-twin : Sky War. This software simulates most of the game's logic and let you test and tune quickly your favourite rush strategy. It also provides a genetic algorithm to polish your strategy : should I create one or two fields ? When should I create them ? The algorithm tries various mutations of your strategie, changing only what it's allowed to, and gives you the mutated strategy with the best fitness.

Unreal 3 Raving Rabbits

I did this during my PhD. The idea was both to learn unreal engine 3 and to create a synthetic player for a custom map. As I am working on difficulty analysis within video games, I wanted to use this synthetic player to wander through a level and find out which part of it was too difficult, or wether or not he could find what to do next.... But this was not the correct approach. Unreal engine is a very good game engine but it's AI is limited to basic state of the art Game AI, using simple path nodes at the lowest level. This is not enough to simulate a real player's behaviour. Another problem was that I had to think about theoretical issues before implementing something worthfull in a game engine, and that if prototyping in unreal engine is quite quick, it's not as quick as coding it with some sprites and OpenGL. So this project ended very soon, but however made me learn how to put a 3ds model into unreal engine, create a level, make custom materials and tune AI scripts. I put it there just for fun.

Game Design Courses

I have made these slides for the Fondation 93 "Innovation" project. Five teachers of my faculty went to five schools around Paris (Saint Denis), and were given 4 courses (2 hours each) to make a game with their class. Luckily, I was one of them, and had the pleasure to teach game design, level design and game maker. This year (2009) is the second edition of this project. Courses are in French.

Happy new year !

We made this video for 2009 happy new year. My first experience with stop motion using flash... not the best choice to sync sound and images, but it worked anyway ! I only took the pictures and edited the movie, the whole design is by Helene Manche, and music is by Jacky Delance.

Solving Snake Cubes

This christmas I got a snake cube. They wanted a tough one so they bought the 4*4*4 snake cube (each side is 4 cubes long). I've broken my brain on it but I can't solve it. So I tried to cheat and found a solution on a website and asked for help at the lab but, even with the solution, no one made it. So as I was really frustrated (and I would really like to talk about difficulty scaling and how not to frustrate the player with the guy responsible for this) I decided to make my computer work for me. I wrote a solver, and tried to made it "graphical" so that I could make something fun out of that pieces a wood. It solves the 4*4*4 in 10 hours, and manages the 3*3*3, much more popular version, in 100ms. I made a slow motion movie, so that I can watch the 3*3*3 cube getting solved. You will find it in that section.

PostScript 3D Renderer

Definitely useless, except if you want to render 3D meshes with your printer. Takes a mesh, a cam pos, and render it with Phong shading. But it was fun coding it in postscript anyway. Another useless geek stuff, thus priceless. Install ghostscript if you want to use it, and run main.ps

Pacman AI

This is an experiment using reinforcement learning and Markov Decision Process to create a pacman AI. I made a c++ clone of pacman, and implemented this AI on top of it. The synthetic player has a partial view of the game. It knows five parameters. The first one had four values, giving the direction of the nearest pellet. The four other dimensions describe the four Pacman directions. For each direction, Pacman knows whether he is facing a wall, a ghost one or two step away from him, a free-of-ghosts path less than 18 steps long, or free-of-ghosts path longer than 18 steps long. You'll find a video of this experiment in that section.

Cocktail Machine

My first experiment with neural networks. I wanted to make a software that learns your taste and generate cocktails for you. I quickly realized that machine learning wasn't magic, and that having a big neural network with many entries equals a huge sample databasis... Anyway I had fun testing generated cocktails and you may still use it as a cocktail database manager !