BDI Architecture

There are several approaches that propose different types of mental attitudes and their relationships. Among them, the most adopted is the belief-desire-intention (BDI) model, originally proposed by Bratman (Bratman 1987) as a philosophical theory of the practical reasoning, explaining the human reasoning with the following attitudes: beliefs, desires and intentions. The essential assumption of the BDI model is that actions are derived from a process named practical reasoning, which is composed of two steps. In the first step, deliberation (of goals), a set of desires is selected to be achieved, according to the current situation of the agent’s beliefs. The second step is responsible for the determination of how these concrete goals produced as a result of the previous step can be achieved by means of the available options for the agent (Wooldridge 2000).

The three mental attitudes that are part of the BDI model are described next.

  • Beliefs. They represent environment characteristics, which are updated accordingly after the perception of each action. They can be seen as the informative component of the system.
  • Desires. They store the information of the goals to be achieved, as well as properties and costs associated with each goal. They represent the motivational state of the system.
  • Intentions. They represent the current action plan chosen. They capture the deliberative component of the system.

Rao& Georgeff (Rao & Georgeff 1995) adopted the BDI model for software agents and presented a formal theory and an abstract BDI interpreter, which is the base for almost all BDI systems, either historical or used at the present. The interpreter operates over beliefs, goals and plans of the agent, which represent the concepts of the mentalistic notions, with small modifications. The most significant change is that goals are a set of consistent concrete desires that can be achieved all together, avoiding the need of a complex phase of goal deliberation. The main task of the interpreter is the realization of the means-end process by means of the selection and execution of plans for a certain goal or event. The first system implemented with success based on this interpreter was the Procedural Reasoning Systems (PRS) (Georgeff & Lansky 1986), which has as a successor the system named dMARS (d’Inverno, Kinny, Luck & Wooldridge 1997, D’Inverno, Luck, Georgeff, Kinny & Wooldridge 2004).

The process of practical reasoning in a BDI agent is presented in the figure below. As shown in this figure, there are seven main components in a BDI agent:


  • a set of current beliefs, representing information the agent has about its current environment;
  • a belief revision function, which takes a perceptual input and the agent’s current beliefs, and on the basis of these, determines a new set of beliefs;
  • an option generation function, (options), which determines the options available to the agent (its desires), on the basis of its current beliefs about its environment and its current intentions;
  • a set of current options, representing possible courses of actions available to the agent;
  • a filter function (filter), which represents the agent’s deliberation process, and which determines the agent’s intentions on the basis of its current beliefs, desires, and intentions;
  • a set of current intentions, representing the agent’s current focus – those states of affairs that it has committed to trying to bring about;
  • an action selection function, which determines an action to perform on the basis of current intentions.


Bratman, M. E. (1987), Intention, Plans, and Practical Reason, Cambridge, MA.
d’Inverno, M., Kinny, D., Luck, M. & Wooldridge, M. (1997), A formal specification of dMARS, in ‘Agent Theories, Architectures, and Languages’, pp. 155-176.
D’Inverno, M., Luck, M., Georgeff, M. P., Kinny, D. & Wooldridge, M. J. (2004), ‘The dMARS architecture: A specification of the distributed multi-agent reasoning system’, Autonomous Agents & Multi-Agent Systems 9, 5-53.
Georgeff, M. & Lansky, A. (1986), Procedural knowledge, in ‘IEEE Special Issue on Knowledge Representation’, Vol. 74, pp. 1383-1398.
Georgeff, M., Pell, B., Pollack, M., Tambe, M. & Wooldridge, M. (1999), The belief-desireintention model of agency, in J. Muller, M. P. Singh & A. S. Rao, eds, ‘Proceedings of the 5th International Workshop on Intelligent Agents V : Agent Theories, Architectures, and Languages (ATAL-98)’, Vol. 1555, Springer-Verlag: Heidelberg, Germany, pp. 1-10.
Rao, A. S. & Georgeff, M. P. (1995), BDI-agents: from theory to practice, in ‘Proceedings of the First Intl. Conference on Multiagent Systems’, San Francisco
Wooldridge, M. (1999), Intelligent agents, in ‘Multiagent systems: a modern approach to distributed artificial intelligence’, MIT Press, Cambridge, MA, USA, pp. 27-77.
Wooldridge, M. J. (2000), Reasoning about Rational Agents, MIT Press.

Leave a Reply