X hits on this document

PDF document

Enhanced NPC Behaviour using Goal Oriented Action Planning - page 25 / 110





25 / 110

2.4 Statement of Research Question

The question being posed by this dissertation was to evaluate whether GOAP is a superior AI technology to FSMs. To answer this question, it was necessary to analyse the two systems on several levels of comparison. The first comparison was to develop the two AI systems and pit them against one another in a game scenario i.e. a Domination game and determine which system performs better. This also includes monitoring how each system can incorporate squad tactics, just like the HTN project (Munóz-Avila and Lee-Urban, 2005). The inclusion of squad tactics helps to highlight the flexibility of each system along with how well the AI system can integrate squad behaviour. Technical tests form the basis of the second comparison, i.e. what are the differences in memory usage, frame rate, CPU utilisation etc. between the two systems. The re-usability, flexibility, ease of management and issues faced during development are the final criteria by which the two systems are evaluated.

The advantages provided by GOAP or AI planning in general have been outlined by several authors but none have performed an in-depth comparison between GOAP and another AI technology to prove these claims or come up with any direct comparisons. Orkin stated the benefits of the GOAP system developed for F.E.A.R. without actually comparing with any other system. Munóz-Avila builds upon GOAP by using HTNs to encode team strategies whilst leaving individual UT bot behaviour to be determined by FSMs and pits them against improved UT bots. So although the work undertaken was similar to these other projects, none actually perform any direct comparisons using GOAP. There has been relatively little work involving GOAP in games, especially for research purposes. Hence this dissertation adds to a knowledge gap in the field of AI planning in games and provides a stepping stone for further work in this area of research.

As FSMs have been so prevalent in games and are a known and stable technology, moving away from them to such a new and unproven technology may not be so appealing for games companies. However the findings from this project not only contribute knowledge


Document info
Document views350
Page views354
Page last viewedTue Jan 17 09:27:30 UTC 2017