Plan patterns for declarative goals in AgentSpeak
Jomi Fred Hübner, Rafael H. Bordini, Michael Wooldridge · 2006
In this paper, we consider the use of declarative goals in the AgentSpeak programming language, originally introduced in [6]. By a declarative goal, we mean a goal that explicitly represents a state of affairs to be achieved, in the sense that, if an agent has a goal p(t1,...,tn) then it expects to eventually believe p(t1,...,tn), and only then can the goal be considered achieved. Currently, although goals form a central component of AgentSpeak programming, they are only implicit in the plans defined by the agent programmer. For example, there is no explicit way of expressing that a goal should be maintained until a certain condition holds; such temporal goal structures are hidden within the plan structures themselves. While one possibility would be to extend the language to introduce an explicit construct for declarative goals and a goal base, we show that this is unnecessary. We introduce a number of plan patterns, corresponding to common types of explicit emporal/conditional (declarative) goal structures, and show how these can be mapped into AgentSpeak code. Thus, a programmer or designer can conceive of a goal at the declarative level, and this goal will be "translated" via these patterns into standard AgentSpeak code.