Agent programming in dribble: from beliefs to goals using plans

Birna van Riemsdijk, Wiebe van der Hoek, John-Jules Ch. Meyer

Abstract: To support the practical development of intelligent agents, several programming languages have been introduced that incorporate concepts from agent logics: on the one hand, we have languages that incorporate beliefs and plans (i.e., procedural goals), and on the other hand, languages that implement the concepts of beliefs and (declarative) goals. We propose the agent programming language Dribble, in which these features of procedural and declarative goals are combined. The language Dribble thus incorporates beliefs and goals as well as planning features. The idea is, that a Dribble agent should be able to select a plan to reach a goal from where it is at a certain point in time. In order to do that, the agent has beliefs, goals and rules to select plans and to create and modify plans. Dribble comes with a formally defined operational semantics and, on top of this semantics, a dynamic logic is constructed that can be used to specify and verify properties of Dribble agents. The correspondence between the logic and the operational semantics is established.