engleski

A knowledge-based system for heuristic search in a competitive multiagent game

In this paper we shall consider possibilities of the heuristic development of a tree of actions and states in a competitive multiagent system. These heuristics are developed on the basis of the agents’ knowledge about the current state of the system and their knowledge concerning the knowledge of other competitive agents. Furthermore, they are based upon deductive conclusions which can be derived by using the available information about the system. The model is based on a simple card game with incomplete information, which means that the developed heuristics actually represent possible strategies used by the player in order to achieve the highest possible win probability. Along with theoretical assumptions, the paper describes a concrete system built in the deductive system Coq for two purposes: the verification of the possibility of implementing the theoretical idea, as well as the statistical measurement of the results achieved by different strategies at different levels of reasoning. For the purpose of reasoning about knowledge, special knowledge matrices have been introduced, which can be used on different levels of nested knowledge. They can also be upgraded on the basis of the available information about the state of the system. They could be used by the players as a source of information for using the chosen strategies.

Knowledge-based system; Competitive multiagent game; Heuristic search; Card game; Coq

nije evidentirano