GOBT: A Synergistic Approach to Game AI Using Goal-Oriented and Utility-Based Planning in Behavior Trees

Abstract

In this paper, we propose a novel game AI framework, the Goal-Oriented Behavior Tree, using Unity game engine for simulations. This framework integrates the advantages of the Goal-Oriented Action Planning architecture and Utility Theory with traditional Behavior Trees, enabling more flexible agent responses to various situations. The simulated environment contains customizable game characters capable of actions like patrol, attack, retreat etc. GOBT allows developers to design agent decision-making processes by applying logic in traditional BTs and using the dynamic planning capabilities of GOAP and utility-based action selection when necessary. The performance of GOBT framework is verified through simulations using a synthetic dataset of agent behaviors in response to changing environmental factors.