Formal models of multi-agent systems

[Chris Landauer]

manor asked about CCS, CSP, and the pi-calculus -
        they are all formal models of interaction among entities -
        i once wrote an implementation of csp compromised towards simulation -
        i use all of them as very good abstract models of interactive processes
        (for modeling and even implementing communication protocols, for 
example) -
        some questions can be answered analytically using these formalisms,
        others seem to be much harder
        (like whether or not the error handling mechanism is adequate)

manor asked about demonstrating the inadequacies of a formalism -
        see below

manor also asked:
        Is there a simple question/problem which you could define, such that
        it has some non-trivial property which should be derivable formally
        (i.e., without "running" it) that would serve as a good benchmark?

my answer is "no" -
        the trouble is (as with many formalisms) that whatever question one 
chooses,
        some formalisms make it easy to answer, others make it hard -
                the problem of when a formalism helps with analytic questions
                is very viewpoint- and question- dependent -
        some things are easy to prove about multiple agent systems
        (for example, it is usually easy to determine whether the number of 
agents
                grows indefinitely or remains bounded),
        other things are much harder
        (for example, is the asymptotic number of agents zero or not) -

for my purposes, if a question is easier to answer by "running" the simulation,
then the formalism is lacking in analytic power,
and if it is easier to answer by some kind of mathematical analysis,
then it (the formalism) is strong enough (for some questions)

more later,
cal


Dr. Christopher Landauer
National Systems Group, The Aerospace Corporation
The Hallmark Building, Suite 187
13873 Park Center Road, Herndon, Virginia 20171
e-mail: address@hidden
Phone: (703) 318-1666, FAX: (703) 318-5409


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================