Jose Maria Novaes dos Santos and Joao Jose Neto, Universidade de Sao Paulo (USP), Brazil

There are formalisms in literature where behavior can be described by a finite set of rules that maps the current device’s state into a new one like the finite state machines, statecharts and petri nets. Those formalisms are named as rule-driven devices. A formal device is said to be adaptive if its behavior changes dynamically in response just to its input stimuli and its current state without any external help. Adaptive rule-driven devices can be used for modeling complex problems in artificial intelligence, natural language, reactive systems, synchronous system and others applications. General game playing (GGP) is a research subfield of Artificial Intelligence which aims at developing systems able to play a variety of games, knowing and understanding the rules only in execution time. Most of the proposed GGP systems are based on statistic methods. This paper aims at a new approach in GGP systems, proposing a solution based on adaptive technology modeling. In brief, a GGP system has been developed in a way that it changes its behavior dynamically in response to rules information and its history of games played. This proposal is presented here as well as results of some games played comparing their strategies approach.

Modelling, Reactive System, Cooperating Adaptive Devices, Rule-Driven Formalisms, Self-Modifying Machine, Adaptive Automata, General Game Playing, Adaptive Device