Post by Moderator: Paul Fishwi » Wed, 02 May 1990 14:57:00

Volume: 2, Issue: 8, Wed May 25 09:57:20 EDT 1988


(1) Simulating Plans
(2) Chaotic Behavior: The Devil's Staircase
(3) Query: Info. on CM and TW

Moderator: Paul Fishwick, Univ. of Florida


Newsgroups: comp.simulation
Date: 23 May 88 19:01:00 GMT

Organization: Advanced Decision Systems, Mt. View, CA (415) 941-3912

I'd like to know what techniques people have used to address the
following problem:

        I'm trying to project future state in a wargame by simulating
a military plan, and playing both sides automatically.  The purpose is
to predict deviations from expected outcomes.  The issue is that the
plan dictates a loosely ordered set of desired actions, but the tokens
on the board also need to respond to the situation at hand.  This
creates a tension in the mechanism used to select actions which I need
to resolve.

        For example, if your plan says to attack, but the situation
pushes you onto the defensive, planned actions either have to be
modified/refined, or held in abeyance.  If the situation has deviated
sufficiently far from expectation, then the plan will have to be
abandoned entirely.  

        The problem boils down to a need for techniques that provide
some reactivity in the execution of a plan.  Many simulation systems
will have had to address this question (though I have only seen a few
relatively simple techniques).

        What approaches have people used?


Subject: Devil's staircase?

Date: Tue, 24 May 88 10:07:44 EDT

  Eugene D. Brooks III  writes:
  If the processors start with a random unconstrained address and start
  a vector request of a common stride (with caviats for say a stride of
  2) the network falls into lock step after a "settling time" with a
  strange periodic behaviour of the memory subsystem.  If a disturbance
  is made in the network after it falls into lockstep it bubbles out the
  conflict and falls into a different periodic motion.

I have been studying asynchronous computations on the Butterfly. I
have been interested in the issue of the dynamics of asynchronous
parallel computation. On possibility that I have thought about, and
you seem to provide evidence of, is the emergence of the so-called
"Devil's staircase" in the dynamics of parallel computation. The
Devil's staircase is the mode-locking/mode-hopping behavior on gets in
non-linearly coupled oscillators. It is described in "The Devils
Staircase", Per Bak, Physics Today. December 1986,38-45. If one plots
the frequency ratio (of two systems; typically a driver oscillator, and
a driven one), against time, one gets a fractal staircase of the
residence times of the modes. Perhaps you can generate plots like this?
For more than two computational entities involved, one would get
fractal terraces and hyperterraces.

Albert Boulanger
BBN Labs


Date: Wed, 25 May 88 01:23:22 EDT
Subject: simulation reference request

Can you refer me to texts describing the two simulation strategies
mentioned frequently in comp.simulation: chandy-misra and time-warp?

                                Thanks. --Guy Hillyer