Belief-desire-intention (BDI) model is one of the most promising approaches dedicated to develop intentional agents. It provides developers with required robustness to deal with a wide variety of complex systems where flexibility and intelligence behavior of agents are needed. Existing BDI architectures are typically composed of a set of customizable components that are composed in a reasoning cycle. The plan selection process is one of the customizable components that has an important influence on the overall agent performance. It consists of selecting among a set of applicable plans the most appropriate one to achieve a given goal. In this paper we propose a generic predictive-based plan selection approach, that aims to select for a given goal, the applicable plan that has the highest potential of being executed successfully. Based on what we refer to as selection indices, it calculates for each plan a success probability that will be used later by the selection process to select the appropriate plan.
展开▼
