Multi-agent Systems

[A]n agent is a … system, situated in some environment, that is capable of flexible autonomous action in order to meet its design objectives. [1]

Jennings et al's restriction of limiting agents to software has been removed (…) here, as Jenning's agents are more properly described as software agents. However, given this change, the Jennings et al requirements of flexibility, autonomous and action to define an agent are relevent to all agents, including humans when acting as agents.

It would also be interesting to consider the relationship of agents to the concept of agency. Some would limit agents to only acting on other's behalf, although perhaps also having motives and goals of their own.

Jennings et al go on to identify the characteristics of multi-agent systems (MAS) as:

  • Agents with incomplete information or capabilities (or resources?) to solve the problem alone;
  • No global system control;
  • Data is decentralised;
  • Computation is asynchonous.

The motivation for MAS systems include robustness and efficency, and solving problems where data, expertise and/or control is distributed.

As well as obvious questions on communications, MAS system design has generated concepts regarding:

  • Joint intention
  • Shared plan modelling
  • Commitment
  • Models for teamwork

The latter point includes questions of building and maintaining multi-party teams [2].

An important factor in co-ordination of agents is trust [3].


Service Science Analysis

Perhaps service systems — the systems that provide services — should be considered as agents, i.e. with incomplete information and resources, to be controlled only through interactions, to have its own data and not to see other's data except through interactions, and to operate asynchonously to other service systems.

This assumption would mean that service co-ordination should be considered in the light of work on multi-agent co-ordination.

Another consideration is that "services" are a subset of the interactions between agents. All services are interactions between agents. The question is whether other interactions exist.


See Also


Related Material

  • Anumba, C.J., Ugwu, O.O. & Ren, Z. (eds) 2005, Agents and Multi-agent Systems in Construction, Taylor & Francis, London.

References
1. Jennings, N.R., Sycara, K. & Woolridge, M. 1998, 'A roadmap of agent research and development', Autonomous Agents and Multi-Agent Systems, vol. 1, pp. 7-38.
2. Tambe, M. & Zhang, W. 1998, 'Towards flexible teamwork in persistent teams', paper presented to the International Conference on Multi-Agent Systems.
3. Castelfranchi, C. & Falcone, R. 1998, 'Principles of trust for MAS: Cognitive anatomy, social importance, and quantification', paper presented to the International Conference on Multi-Agent Systems.
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License