Autoplay: Automatic player creation using conceptual clustering

Date of Publication


Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Computer Science

Subject Categories

Computer Sciences


College of Computer Studies


Computer Science

Thesis Adviser

Merlin Teodosio C. Suarez

Defense Panel Member

Danny Cheng

Rafael A. Cabredo


Real-Time Strategy (RTS) games entail a lot of difficulty due to the sheer number of tasks to perform (i.e. attach, build) and factors to consider (i.e strategy-formulation, resource-handling). However, the main source of challenge found in RTS games comes from an opponent. A limitation exists in that current AI agents in games need to be given an unfair advantage to simulate difficulty. Once a payer learns to counter the strategy, it may lead to a decrease in the variety, difficulty, and playability of a game if he cannot find human opponents. An agent capable of learning and modeling a player's moves can help alleviate this problem by providing a means for the agent to vary its moves from a prewritten, scripted AI. User modeling (UM) is the process of obtaining relevant user information to be able to create a model based from a user's behavior (Rosson, 1998). This research applies UM by having an agent model a player and uses this to implement a capable AI opponent. This is done by obtaining relevant user actions and corresponding environment states. This is done by obtaining relevant user actions and corresponding environment states. This research tackles problems involve in modifying an RTS game to fit the agent (for the player to train or fight) and the user model, and constructing a user model. All these three algorithms failed to come up with a useful model. The set of games was then trimmed down to 7 games, and the algorithms constructed the user model of a player and the agent acted like the player. However, there were several factors that greatly hindered the agent's performance. Some of these factors were the inclusion of irrelevant attributes and failing to consider relevant attributes.

Abstract Format






Accession Number


Shelf Location

Archives, The Learning Commons, 12F, Henry Sy Sr. Hall

This document is currently not available here.