Toward Obstacle Avoidance for FPS Bots

Student thesis: Master thesis (including HD thesis)

  • Mikkel Mølgaard Madsen
4. term, Software, Master (Master Programme)
First-Person Shooter (FPS) is a genre of shooting-centric computer
games where the player controls a character in the game, and sees the
virtual world through that character’s eyes. FPS games often include
AI opponents in some form, and the game world very often contains
different objects and terrain, much of which can be considered obstacles
for an AI. This thesis attempts to take the first steps toward a general-
purpose solution for an AI module which can learn based on samples
from human players, how to navigate around obstacles. Two machine
learning techniques are implemented and compared for this purpose -
Support Vector Machines and Elman Recurrent Neural Network.
Publication date7 Jun 2013
Number of pages60
ID: 77337219