Angry Birds is a video game created by Rovio Entertainment in 2009. The game was immediately pretty successful and popular, 12 million copies of the original game had been sold in the first year. The goal is to destroy pigs protected by a structure composed of different building blocks that observe the laws of physics. The structure can be destroyed or displaced by shooting the angry birds at it. The fewer birds used and the more blocks destroyed, the higher the score will be. Due to the nature of this game (large state and action spaces, various object properties, noisy object detection, not deterministic physical model), Angry Birds constitutes a really challenging task for AI. The idea of this thesis comes from a competition, namely the Angry Birds AI Competition aibirds which is based on the video game port on chrome. The competition has been organized each year since 2012 when the very first Angry Birds AI competition was held in Sydney, as part of the 25th Australian AI conference. There is a framework for the agents developed for the competition, which includes Google Chrome plugins allowing our agent to interact with the game. Other available resources include a Naive Agent and the base framework as explained below. The goal of our thesis is to develop an interesting agent that can play the game and score well relative to the benchmarks from previous years of the AIBirds competition. We will start by a description of the Angry Birds game we consider, explaining the goals, all the tools we have at our disposal and the mechanics of the game. A state of the art section, explaining some of the most successful agents submitted for the competition follows. Then, we will discuss the first serious heuristic agent we created. We then follow up with the more complex machine learning based agent. Our agents are compared, evaluated and we explore a few improvement ideas in the subsequent sections. ; Master [120] : ingénieur civil en informatique, Université catholique de Louvain, 2016 ; Master [60] en sciences informatiques, Université catholique de Louvain, 2016


    Access

    Download


    Export, share and cite



    Title :

    AI for games : Angry Birds



    Publication date :

    2016-01-01


    Type of media :

    Theses


    Type of material :

    Electronic Resource


    Language :

    English



    Classification :

    DDC:    629



    AI for games: Angry Birds.

    Hollevoet, Pieter / Bertaux, Jérôme | BASE | 2016

    Free access

    WHY SHIPPERS ARE ANGRY

    Online Contents | 1997


    Islands of angry ghosts

    Edwards, Hugh | SLUB | 1966