Strategic Cooperation in First-Person Shooters

Studenteropgave: Kandidatspeciale og HD afgangsprojekt

  • Alex Vendelbo Ringgaard
  • Martin Guld Thomsen
4. semester, Datalogi, Kandidat (Kandidatuddannelse)
This report presents two agent centered techniques for cooperation among agents. I.e. techniques to support cooperation among agents, without the need for the agents to engage in negotiation. This is for example needed to cooperate with human players in fast paced games like first-person shooters, where the human players only have time to send short messages to the agent. The two techniques presented in this report is: Prediction, where an agent attempt to determine what actions other agents has or will take. This technique is used as the primary way to support cooperation, as it does not rely on any communication. The second technique, called value of communication, is used to determine whether it is worth to attempt to communicate in a given situation. This means both requesting information from another agent, as well as answering a request sent by another agent. This technique is used as a secondary technique to support the prediction, as it does rely on some message passing between agents. While both techniques are general enough to apply to most areas of cooperation, this report focuses on strategical cooperation among defending agents in the game ''''Team Fortress'''' to show, through an empirical evaluation, that the techniques improve an agent''s ability to achieve its goal. The report concludes that the prediction technique is successful in improving the capabilities of the agents, and that the value of communication technique is capable of determining the most profitable times to request information from other agents.
Udgivelsesdatojul. 2004
ID: 61061617