A Framework for Defining Intelligent Agents with Different Orientations that take Decisions based on a Planning Graph in a Game Environment

Abstract

This dissertation proposes a framework for the development of intelligent agents in Roleplaying games where planning graphs are used to allocate actions to the agents and they choose to perform these actions based on their orientation. The abilities contained in them help in defining their orientation as well as their ability to perform actions during decision making activity. The overall goal of this research is to generate a different animated story every time the system is run. By distributing a game environment into a collection of characters and items and using actions as the mode of interaction between the two, this random story generation is achievable. The physicality of the characters is dependent on the choices made by the users in previous runs of the system. The system adapts to these choices and generates a character based on them. This ensures that if this system is run at different areas around the world, it will start to create characters that are more according to the likes of the people in that part of the world. This required a system that could remember the users’ choices and evolve with increased user input. For this reason two separate genetic algorithms were used to define the orientation of the characters through the allocation of abilities and for defining the physical characteristics. For allocating the right abilities in a character’s chromosome a genetic algorithm is used to solve this knapsack problem. For deciding the physical characteristics of characters, genetic algorithm with sigma selection is used to narrow down the choice for the most likable physical appearance based on previous user choices. A story is dependent on the actions done by the agents; if we string these actions together, we can generate a meaningful story. Actions are dependent on the goals that need to be achieved by the characters. These actions need to be in a certain order to allow efficient completion of goals, which means that every character must formulate a