Modelling BDI-based Agents in a Real-time Action Game

From Master Projects
Jump to: navigation, search


has title::Modelling BDI-based Agents in a Real-time Action Game
status: finished
Master: project within::Technical Artificial Intelligence
Student name: student name::Jeremy Soumokil
number: student number::1204475
Dates
Start start date:=2009/02/02
End end date:=2009/08/02
Supervision
Supervisor: Mark Hoogendoorn
Second reader: has second reader::Tibor Bosse
Company: has company::W!Games
Poster: has poster::Media:Media-poster.pdf

Signature supervisor



..................................

Abstract

KIM 1: Abstract

Presentation date: April 24, 2009 in M648

In search for a solution for modeling proactive behavior of non-player characters in a real-time action game, the Belief-Desire-Intention (BDI) model has promising theoretical foundations. BDI adheres to the required characteristics of the goal-based agent architecture. The most successful control architecture for this model is the Procedural Reasoning System (PRS). PRS is a general purpose reasoning system of which many different variations have been devised and applied, i.e. to robotics, air traffic management and server-side customer-service applications.

In this project I will explore the possibilities of applying such BDI approaches within the constraints of a game environment. By iterative designing and implementation I will attempt to create a BDI architecture that produces appropriate agent behavior from reasoning about the actions of the human player. This process gradually exposes advantages, disadvantages and opportunities that BDI has to offer in favor of the commonly used finite state machines and decision trees.

KIM 2: Abstract

Presentation date: August 13, 2009 in S624

One of the main challenges of game AI is handling interaction between in game characters and the player. Making correct decisions requires knowledge about the other’s actions and motives. In this work we investigated how Theory of Mind (ToM) principles can assist in addressing this problem.

The presentation will propose a light-weight implementation of the Procedural Reasoning System (PRS). PRS regulates the control flow of Belief-Desire-Intention (BDI) constructs, which are used to model the behavior of the agent. We show how ToM can be used and how it affects behavioral modeling.