Trained Behavior Trees: Programming by Demonstration to Support AI Game Designers

Trained Behavior Trees: Programming by Demonstration to Support AI Game Designers. Sagredo-Olivenza, I., Gomez-Martin, P., P., Gomez-Martin, M., A., & Gonzalez-Calero, P., A. IEEE Transactions on Games, 11(1):5-14, 2017.

Paper doi abstract bibtex

Programming by demonstration (PbD) has a straightforward application in the development of the artificial intelligence (AI) for non-player characters (NPCs) in a games: a game designer controls the NPC during a training session in the game, and thus demon- strates the expected behavior for that character in dif- ferent situations. Afterwards, applying some machine learning technique on the traces recorded during the demonstration, an AI for the NPC can be generated. Nevertheless, with this approach, it is very hard for the game designer to fully control the resulting behavior, which is a key requirement for game designers, who are responsible for putting together a fun experience for the player. In this paper, we present Trained Behavior Trees (TBTs). TBTs are behavior trees (BTs) generated from traces obtained in a game through programming by demonstration. BTs are a technique widely used for AI programming which are created and modified through special purpose visual editors. By inducing a BTs from a programming by demonstration game session, we combine the ease of use of this technique with the ability to fine-tune the learned behavior of BTs. Fur- thermore, TBTs facilitate the use of behavior trees by game designers and promote their authoring control on game AI

@article{
 title = {Trained Behavior Trees: Programming by Demonstration to Support AI Game Designers},
 type = {article},
 year = {2017},
 pages = {5-14},
 volume = {11},
 id = {3f785d3f-cd57-3ab4-8f61-d778de93e390},
 created = {2019-05-16T09:42:44.590Z},
 file_attached = {true},
 profile_id = {7ff3d559-34c5-3dc7-a15e-4809d39e6685},
 group_id = {42178b72-b294-3d2f-9512-ce9371d745ab},
 last_modified = {2019-05-16T09:42:48.211Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 private_publication = {false},
 abstract = {Programming by demonstration (PbD) has a straightforward application in the development of the artificial intelligence (AI) for non-player characters (NPCs) in a games: a game designer controls the NPC during a training session in the game, and thus demon- strates the expected behavior for that character in dif- ferent situations. Afterwards, applying some machine learning technique on the traces recorded during the demonstration, an AI for the NPC can be generated. Nevertheless, with this approach, it is very hard for the game designer to fully control the resulting behavior, which is a key requirement for game designers, who are responsible for putting together a fun experience for the player. In this paper, we present Trained Behavior Trees (TBTs). TBTs are behavior trees (BTs) generated from traces obtained in a game through programming by demonstration. BTs are a technique widely used for AI programming which are created and modified through special purpose visual editors. By inducing a BTs from a programming by demonstration game session, we combine the ease of use of this technique with the ability to fine-tune the learned behavior of BTs. Fur- thermore, TBTs facilitate the use of behavior trees by game designers and promote their authoring control on game AI},
 bibtype = {article},
 author = {Sagredo-Olivenza, Ismael and Gomez-Martin, Pedro Pablo and Gomez-Martin, Marco Antonio and Gonzalez-Calero, Pedro Antonio},
 doi = {10.1109/tg.2017.2771831},
 journal = {IEEE Transactions on Games},
 number = {1}
}

Downloads: 0