Work at SourceForge, help us to make it a better place! We have an immediate need for a Support Technician in our San Francisco or Denver office.
A characteristic of the BDI model is intention reconsideration. How is it implemented in Jadex?
I'd like to probe different strategies of intention reconsideration. Can I do this?
Thanks in advance.
in BDI there are different perspectives depending on whether you look at BDI from the
a) philosophical (Bratman)
b) logical (Rao)
c) implementation (PRS from Rao+Georgeff, or IRMA from Bratman+Pollack)
At the implementation level, an intention is more or less a plan stack, i.e. a cuurent hierarchy of goals and plan the agent has committed to. As in Jadex goals are not only procedural but also declarative when using conditions, intention reconsideration is performed whenever such conditions change.
If you have something different in mind please let us know.
Hi, thanks for answer me.
Kinny and Georgeff did a study about of the dilemma of balancing proactivo (goal directed) and reactive (event directed) behavior. They investigated bold and cautious agents in different environments (commitment strategies).
Given that changes due to the dynamic characteristics of the environment do not necessarily induce intention reconsideration, I'm researching a mechanism to decide whether deliberate or act based on possible future events , i.e. to use an adaptive strategy.
I believe that Jadex use cautious type intention reconsideration, it always reconsider when a change in the goal'conditions occurs, which is not always necessary. On the other hand, in a non-deterministic environment sometimes could be necessary to reconsider rather than replan (so I think that if not has been a change in the goals, reconsideation doesn't happen in Jadex.)
I know this stude of Kinny and Georgeff but I am not sure if these observations are still that helpful with todays agents. You are right that Jadex agents are cautious in the sense that they immediate act when conditions change, i.e immediately know when a goal is achieved etc. As far as I remember they found out that:
a) the more dynamic an environment is the more cautious agents should be
b) bold agents have advantages wrt to fast execution (as they do not need to replan frequently)
But given that on todays machines hundreds or thousands of agents can be executed without problems what is the rationale of not immediately reacting to changes? Do you have a specific application case in mind?