X hits on this document

PDF document

Enhanced NPC Behaviour using Goal Oriented Action Planning - page 27 / 110





27 / 110

Chapter 3. DESIGN

3.1 The Game Scenario

It was decided from Munóz-Avila’s article (Munóz-Avila and Lee-Urban, 2005) to perform the experiments between the two types of AI in a Domination game for this dissertation. This game is an ideal test scenario for the two types of AI as it puts one team against another and the scores and kills from the game can be easily measured and compared. A certain degree of squad tactics are required for a team to be successful so this also enables the evaluation of how well each type of AI can integrate squad behaviour. HTNs were used to control the squad coordination in Munóz-Avila’s work however the extra complexity they bring meant that it was decided not to implement them for this project. An ad-hoc approach to squad coordination and tactics was designed in its stead, similar to the approach taken by Orkin in designing the squad behaviour for F.E.A.R.. The squad system prioritises agent goals but each agent uses its own individual GOAP system to make decisions.

Ideally, this project would have modded UT2005 to create the Domination game simulation just like the HTN project. However due to the time constraints and man-power available, this wasn’t an option. So instead it was decided to develop a 2D game using Microsoft’s DirectX API that mimics the UT2005 Domination game mode as closely as possible. DirectX is quite mature, widely used in industry and minimized any learning curve involved with making the game. It also allowed more focus on the actual AI rather than having to get to grips with a large codebase and complex systems that would have occurred if the modding of UT2005 was undertaken. The option of extending the Domination game to include other game modes was left open in the design of the game.


Document info
Document views375
Page views379
Page last viewedThu Jan 19 22:09:12 UTC 2017