DocumentCode :
2217536
Title :
Financial control of the evolution of autonomous non-player characters
Author :
Ashlock, Daniel ; Nguyen, Sylvia
Author_Institution :
Dept. of Math & Stat, Univ. of Guelph, Guelph, ON, Canada
fYear :
2011
fDate :
5-8 June 2011
Firstpage :
828
Lastpage :
835
Abstract :
This study prototypes a method of evolving autonomous agents that can act as non-player characters (NPCs) in a game. The agents move based on information about their local environment and have evolved weapons, armor, ability to take damage, and movement factors. The creation of the agent is divided into two phases. In the first, a population of competent movement controllers are evolved. In the second, agents start with a competent movement controller and evolve weapons, levels of armor, number of hitpoints, and numbers of movement factors. The movement controller continues to evolve in the second phase. The evolution of the agent\´s equipment is constrained by a budget together with a price for each type of object the agent can have. The gene specifying the agent\´s equipment is in the form of a "wish list" of equipment, traversed left-to-right, with the agent buying items from the list as long as its budget suffices. A agent that is a more dangerous opponent can be evolved by giving it a larger budget. A group of experiment are performed that demonstrate that the budget can be used to control an agent\´s toughness. Additional experiments show that changing the price list for different items can also be used to control the types of agents that evolve. Pitfalls in the selection of the fitness function for the agents are discussed.
Keywords :
artificial intelligence; budgeting; computer games; armor; autonomous agent; autonomous nonplayer characters; competent movement controllers; evolved weapons; financial control; fitness function; game; movement factors; Arrays; Blades; Economics; Games; Mathematical model; Robots; Weapons;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Evolutionary Computation (CEC), 2011 IEEE Congress on
Conference_Location :
New Orleans, LA
ISSN :
Pending
Print_ISBN :
978-1-4244-7834-7
Type :
conf
DOI :
10.1109/CEC.2011.5949704
Filename :
5949704
Link To Document :
بازگشت