[Zurück]

J. Inführ, G. Raidl:
"Automatic generation of 2-AntWars players with Genetic Programming";
Vortrag: International Conference on Computer Aided Systems Theory (Eurocast), Gran Canaria, Spain; 06.02.2011 - 11.02.2011; in: "Extended Abstracts of EUROCAST 2011 - 13th International Conference on Computer Aided Systems Theory", (2011), ISBN: 978-84-693-9560-8; S. 244 - 246.

AntWars is a competitive two-player game with local information that was introduced
as part of a competition accompanying the Genetic and Evolutionary
Computation Conference 2007 [1,3]. Both players control an ant in a toroidal
world and have to collect randomly placed pieces of food. The player who collects
more food wins.
2-AntWars is an extension of AntWars. In 2-AntWars, each player controls
two ants in a rectangular world four times the size of the AntWars world. Controlling
two ants increases the complexity of the problem considerably because
now each player has to decide which ant to move in addition to selecting the
direction of the move, and he has to keep the location of the ants in mind because
moving an ant into the boundary of the world would make it immovable.
Furthermore, the decision to battle with an ant of the opponent (by moving an
ant to a location that is occupied by an ant of the enemy) requires more nesse
than in AntWars. In AntWars, the aggressor wins instantly as the player who is
attacked can not counteract. In 2-AntWars, the defending player has the possibility
to move his second ant to the position of the battle to win. The complete
description of 2-AntWars can be found in [2].
In this work we studied how Genetic Programming can be used to create
competent 2-AntWars players.

http://publik.tuwien.ac.at/files/PubDat_201099.pdf