Information Technology Reference
In-Depth Information
14.5.7
Intention based negotiation
Grosz applied the belief, desire and intention theory to the agent negotiation
(Grosz et al,1996). This method does not use the sub plan, while use the intention
for negotiation to decrease the communication. BDI theory thinks that the agents'
actions were not the desire and plan but the result of combination of belief and
desire. The sub plan to realize the intension is generated by this intension. One
intension is correspondence with some sub plans. Agents' communication need
not to exchange the sub plans and only exchange the intentions. However,
Grosz's approach assumes the agents are completely cooperative.
Zlotkin's work greatly improves the static negotiation theory of two agents.
However, to open multi-agent system, his approach is not applicable. Current
BDI team approaches lack tools for quantitative performance analysis under
uncertainty. Distributed partially observable Markov decision problems
(POMDPs) are well suited for such analysis, but the complexity of finding
optimal policies in such models is highly intractable. Nair and Tambe have
proposed a hybrid BDI-POMDP approach, where BDI team plans are exploited
to improve POMDP tractability and POMDP analysis improves BDI team plan
performance (Nair et al,2005).
14.5.8
Team-oriented collaboration
Coordination between large teams of highly heterogeneous entities will change
the way complex goals are pursued in real world environments. Scerri and his
colleagues proposed a Machinetta approach which combines Team Oriented
Programming and proxy architecture to overcome the limitations of effective
coordination between very large teams of highly heterogeneous agents (Scerri,
2003). The main advantages to Machinetta approach display one or a
combination of the characteristics: large scale, dynamic environment, and
integration of humans. By connecting the Machinetta proxies with the graphical
development tool for constructing team plans, the Team Oriented Programming
programmer gains a good idea of what is going on in the plan and how to make
effective changes in it.
Token-based coordination is a process by which agents attempt to maximize
the overall team reward by moving tokens around the team. If an agent were to
know the exact state of the team, it could use an Markov Decision Process (MDP)
to determine the expected utility maximizing way to move tokens. Unfortunately,
it is infeasible for an agent to know the complete state, however, it is illustrative
