ReAd: reactive-adaptive methodology to enable evolving intelligent agents for virtual environments

Abstract

This paper describes our novel reactive-adaptive methodology (ReAd) for the creation of Intelligent Agents capable of evolving to self-develop, in virtual environments. We start with AI concepts, which are well established for the implementation of character behaviour in serious games, such as Fuzzy Logic, the Belief-DesireIntention model (BDI), and Finite State Machines (FSM); and discuss their characteristics. In particular for BDI and FSM, we analyse their limitations for being manipulated at run-time, which in turn limits their use in evolvable systems. We present a novel combination of these techniques, based on a Rational-Reactive structure (RaRe) to optimize their performance and enable the process of online selfadaptation so that they can be used to create evolving intelligent agents. The focus of the work is in enabling a structure to be evolvable; the detail of the adaptation process itself is not in the critical domain of this paper. We present an analysis of our system in a test scenario, where the standard implementation is compared to our novel ReAd methodology.