Print

Print


Dear Peer:

Concise answers are always insufficient but they have the advantage of marking a relevant difference.

ABM is a bottom-up approach to generate signatures of the system to be analyzed, just specifying the agents endorsements (perhaps obtained from survey research) and the institutional rules.
To simulate-explore-experiment with ABM you have to detail
•    the agent’s endorsements,
•    the environmental assumptions,
•    and the institutional setting.
•   
Take a market. You need to specify the buyers and sellers endorsements (reservation prices) as “autonomous agents”. The actual aggregated supply and demand (the environment, where you randomly picked numbers for the reservation prices to be assigned to the agents) . How the trade is to be conducted (for example the buyers listen the prices put by the sellers can not tell their reservation prices).
That ´s all you need.

In currently used DS simulators, you use a top down approach. You establish the agents network and flows before simulating….in many cases “wishful modeling”, or a procustres approach. You buy the coffin and then if the corpse is too long you just cut the legs, and if it is too short you just stretched it to fit the coffin,,,

Of course, both methods are complementary.
You first use ABM to have systems responses (result of the simulations). Then you analyze the aggregated behavior of the signatures, and you provide a DES version of the model, compatible with these signatures…and go on and on, till you are lucky to find “the true” model.

But you see the essence is to start with an ABM so as to avoid contaminating the model ab initio by the modeler, and to see how much can you grasp about the behavioral aggregated patterns and to what extent your results depend on the agents endorsements, the environment and the institutional setting...(calibration). Yes it sounds a tedious task but this is modeling without too much theory.

Af¡ just a “simple warning” …the agents are not objects.

Have I been of any help?  Probably not.

Best regards

Cesareo


Peer-Olaf Siebers escribió:
[log in to unmask]">
Hello

I'm currently investigating the research question: Can Agent-Based
Simulation (ABS) help us with assessing the impact of human resource
management practices on customer satisfaction and the performance of
service-oriented retail organisations?

I have often been asked the following questions after presenting my work:
- Why do you use ABS?
- Why can't you do it with Discrete Event Simulation (DES)?
- Or do you use the agent approach just as a mindset rather than a technique?

Regarding these questions I never came to a satisfying conclusion! I thought
it would be interesting to collect the different views that people have
about this question, so I would like to discuss it here!

The questions are:
- What is the difference between ABS and DES?
- Why do we use Agent-Based Modelling (ABM) and ABS?

I have the feeling that very often people mix up terminology. Therefore, I
would like to sketch out how I understand the different terms:

There are two simulation technologies:
- Continuous Simulation
- Discrete (Event?) Simulation

There are three different paradigms (are these modelling paradigms?)
- System Dynamics (SD) paradigm
- Discrete Event (DE) paradigm
- Agent Based (AB) paradigm

For human-centred systems, when it is important to consider a heterogeneous
population, the AB paradigm seems to be the most useful approach to use for
the modelling process. There is a structural correspondence between the real
system and the model representation, which makes them more intuitive and
easier to understand.

How do we actually (technically) implement an ABM?

There are three different world views (approaches) for DES. The different
simulation systems provide facilities for model implementation according to
the world view. The three approaches are:
- Activity Scanning
- Event Scheduling (most frequently associated with the phrase DES) 
- Process Interaction

Garrido (2001, p59) defines the Process Interaction World View as follows:
"In the process world view, the behaviour of the system under study is
represented by a set of interacting processes. Recall that a process is
defined as a set of abstract data structures and the sequence of operations
carried out by an entity during its life within the system. The merging of
the event sequences of these processes contains all events that occur in the
system."

For me, this sounds very much like a definition for ABS. So we use the AB
paradigm to develop our conceptual ideas and models (ABM) and then the DES
process interaction approach (which we call ABS) to implement the models. 

Are there other ways to implement ABS models? Or is ABS just a buzz word? Is
it just a consequence that if you want to implement ABMs you want to use an
AB technology, hence a new name for an old method?

Thanks for reading - I'm looking forward to hearing your comments!



If you're interested in the project you can find more information and
publications at http://www.cs.nott.ac.uk/~pos



Definitions:
- A technology is a manner of accomplishing a task especially using
technical processes, methods, or knowledge (Merriam-Webster Online
Dictionary: http://mw4.m-w.com/dictionary/technology).

- A paradigm is a philosophical and theoretical framework of a scientific
school or discipline within which theories, laws, and generalizations and
the experiments performed in support of them are formulated (Merriam-Webster
Online Dictionary: http://mw4.m-w.com/dictionary/paradigm).

- A worldview is a comprehensive conception or apprehension of the world
especially from a specific standpoint (Merriam-Webster Online Dictionary:
http://mw4.m-w.com/dictionary/weltanschauung).



Reference:
- J.M. Garrido (2001). Object-Oriented Discrete-Event Simulation with Java:
A practical introduction. Kluwer Academic: NY

  

--
De: Cesáreo Hernandez Iglesias
Fecha: Thu, 13 Jul 2006 13:34:39 +0200

Cesareo Hernández Iglesias
Depto. de Organización de Empresas y C.I.M.
E.T.S. de Ingenieros Industriales
C/ Paseo del Cauce s/n
47011 Valladolid
Tlfno: +34 983 423336
Fax: +34 983 423310
email: [log in to unmask]