Agent Anatomy

In order to obtain autonomy for the agents within a system various internal anatomies have been proposed depending on the task that they are required to fulfil. Luck et al (1) categorises agent anatomies in three ways: reactive, deliberative and hybrid.

Reactive agents

These agents are reactive in nature and do not hold a symbolic view of their environment. They are based on the subsumption architecture which was first proposed by Brooks (2). In this architecture, a hierarchy of layers is defined by how specific a behaviour is. Each layer, or behaviour, in the architecture is independent of the others, but a higher level behaviour is able to suppress the behaviour of a lower level one. This type of anatomy would be applicable for agents that require no reasoning ability but where fast action, if necessary, could be taken. Drawbacks of this type of anatomy are that the agents are unable to adapt or cope with uncertainty and it may also be difficult to predict an agent’s action for a given situation.

Deliberative agents

Deliberative agents contain an explicitly represented, symbolic model of the world, where actions are decided through reasoning (3). Agents with this type of anatomy generate plans to accomplish their goals and are therefore better suited to coping with uncertainty, reacting to unforeseen circumstances and recovering from poor decisions (4).

Several models have been proposed for building deliberative agents with many on the BDI (Belief, Desire, Intention) model (5). In this model, the agents beliefs represent what the agent knows about itself and its environment. The agent’s desires are its motivations, i.e. what the agent is trying to achieve, but where the agent has multiple desires it is possible that these desires conflict. The intentions of the agent represent the sequence of actions required for the agent to achieve its goals i.e. the agent’s plans.

Other popular models for the anatomy of deliberative agents are model-based reasoning (MBR), notably used by Williams et al (6), and the Procedural Reasoning System (PRS) developed by Georgeff and Lansky (7), the interested reader should also investigate the dMARS architecture (8), which is an extension of the PRS. Usually the main drawback with deliberative agents is their reaction time given a stimulus; some systems, such as the PRS, aim to combat this by using a plan library for common desires which define sequences of low level actions which are triggered by invocation conditions.

Hybrid agents

In practice most agents will be a combination of reactive and deliberative, examples of hybrid anatomies are TouringMachines (9) and INTERRRAP (10).

Tools for developing MAS

Various commercial and open source MAS development tools are available (11); it is JADE (12) that appears to have become most popular. A review of the main toolkits available are detailed in Luck et al (1).

Leigh Tesfatsion maintains a list of open source software resources for market simulation, including the agent-based AMES toolkit.

References

  1. M. Luck, R. Ashri, and M. D’Inverno. Agent-Based Software Development, Artech House Publishers, Boston, 2004.
  2. R. A. Brooks, “Intelligence without representation,” ser. Artificial Intelligence, 1991, no. 47, pp. 139-159.
  3. M. J. Wooldridge and N.R. Jennings, Agent theories, architectures, and languages: a survey, Springer-Verlag, 1995.
  4. I. A. Ferguson, Integrated Control and Coordinated Behaviour: A Case for Agent Models, Proceedings of the 1994 Workshop on Agent Theories, Architectures, and Languages (ATAL 94). Amsterdam, Netherlands. Lecture Notes in Computer Science Series, Springer-Verlag. Berlin, Germany. August 1994.
  5. A. S. Rao and M. P. Georgeff, “BDI agents: From theory to practice,” in First International Conference on Multi-Agent Systems (ICMAS-95), 1995.
  6. M. Georgeff and A. Lansky, “Reactive reasoning and planning,” in AAAI-87 Proceedings, 1987.
  7. B. C. Williams, M. D. Ingham, S. H. Chung and P. H. Elliott, “Model-based programming of intelligent embedded systems and robotic space explorers,” Proc. IEEE, vol. 91, no. 1, pp 212- 237, Jan. 2003.
  8. M. d’Inverno, D. Kinny, M. Luck, and M. Wooldridge, “A formal specification of dMARS,” in Proceedings of the Fourth International Workshop on Agent Theories, 1998.
  9. J. F. Lehman, J. Laird, and P. Rosenbloom, “A gentle introduction to SOAR, an architecture for human cognition,” Invitation to Cognitive Science, vol. 4, 1996.
  10. K. Fischer, J. P. Muller, and M. Pischel, “Unifying control in a layered agent architecture,” in IJCAI95, Agent Theory, Architecture and Language Workshop 95, 1995.
  11. Software development tools for Multi-Agent Systems
  12. Java Agent Development Framework (JADE)

Citation

If you want to reference material on this topic, please consider citing these papers as appropriate:

  • S. D. J. McArthur; E. M. Davidson; V. M. Catterson; A. L. Dimeas; N. D. Hatziargyriou; F. Ponci; T. Funabashi, “Multi-Agent Systems for Power Engineering Applications—Part I: Concepts, Approaches, and Technical Challenges”, IEEE Transactions on Power Systems, Vol. 22, No. 4, November 2007.
  • S. D. J. McArthur; E. M. Davidson; V. M. Catterson; A. L. Dimeas; N. D. Hatziargyriou; F. Ponci; T. Funabashi, “Multi-Agent Systems for Power Engineering Applications—Part II: Technologies, Standards, and Tools for Building Multi-agent Systems”, IEEE Transactions on Power Systems, Vol. 22, No. 4, November 2007.