|
Investigating the management of conflicts in multi-agent
systems is an activity which involves detecting that there is a conflict,
deciding how the conflict can be resolved and then executing the necessary
repair actions. Of particular interest is the use of negotiation techniques
to this problem and their implementation in a distributed truth maintenance
system (TMS). Developing the distributed TMS has involved extending the
standard asocial TMS to deal with beliefs that have originated from other
agents in the system (in addition to those which the agent has developed itself). In such cases a number of crucial decisions must be made about
how the information provided by other agents should be treated - should
it be given the same credence as locally deduced beliefs?, should it only
be used when there is supporting local evidence?, how should contradictions
between the beliefs of different agents be dealt with?, and so on. It is
in the discussion and resolution of these issues and their subsequent implementation
that the main contribution of this work lies. In 1997, the main work was
focussed on the maintenance of the distributed beliefs' consistency in
the specific control scenario proposed by CERN. In 1998, the research carried
out was concentrated on the development of methodologies for solving conflicting
beliefs.
|