Goal Selection Strategies for Rational Agents

In agent theory and agent programming, goals constitute the motivational attitude of rational agents and form the key concept in explaining and generating their pro-active behavior. Pursuing multiple goals simultaneously might pose problems for agents as the plans for achieving them may conflict. We argue that a BDI-based agent programming language should provide constructs to allow an agent programmer to implement agents that: 1) do not pursue goals with conflicting plans simultaneously, and 2) can choose from goals with conflicting plans. This paper presents an explicit and generic mechanism to process incompatible goals, i.e., goals with conflicting plans. The proposed mechanism can be integrated in existing BDI-based agent programming languages. We discuss different strategies to process incompatible goals based on a given conflict relation and show some properties and relations between these strategies.