Implementing fast planning with regards to Game-AI | Dan True, Jakob Holmelund
| | Abstract | In this project we develop and present a formal planning domain with key features alike to a classic action game. Areas covered are dynamic environments, partial-observability, hierarchical move actions, execution monitoring and (re)planning. We analyse a few well-known game genres and establish a set of properties they share with regards to planning. We present existing and propose new theoretical solutions to the problems inherent in these types of domains. We analyse three well-known planning techniques: Forward-State Space (FSP), GraphPlan & Partial-Order Planning(POP) and their application to our specified domain. We then implement an FSP planner and a POP planner and measure their performance in the environment. We develop an agent taking advantage of a number of the proposed solutions. This agent iterates through a queue of goals, receives percepts, maintains a belief base representing the world, formulates and monitors plans and executes actions. To test and demonstrate the agent method we have designed a simple graphical user interface, in which objects can be placed and moved dynamically during runtime. The agent is added to the environment and given goals to fulfil, overcoming problems caused by the dynamic environment, partial-observability etc. during run-time. | | Type | Bachelor thesis [Academic thesis] | | Year | 2012 | | Publisher | Technical University of Denmark, DTU Informatics, E-mail: reception@imm.dtu.dk | | Address | Asmussens Alle, Building 305, DK-2800 Kgs. Lyngby, Denmark | | Series | IMM-B.Sc.-2012-09 | | Note | Supervised by Associate Professor Thomas Bolander, tb@imm.dtu.dk, DTU Informatics | | Electronic version(s) | [pdf] | | Publication link | http://www.imm.dtu.dk/English.aspx | | BibTeX data | [bibtex] | | IMM Group(s) | Computer Science & Engineering |
|