This document describes a software framework that integrates agent-based simulations with system dynamics models. It allows agents in an agent-based simulation to call system dynamics models. The framework uses Java for the agent-based components and interfaces with the system dynamics software Vensim using a DLL. An example application to a supply chain model is provided, where manufacturer agents run a system dynamics inventory control model in each time step. The integration allows leveraging the strengths of both approaches for more flexible modeling of dynamic systems.

1. A Software Interface
Between System
Dynamics and Agent-
Based Simulations


2. A study by
Andreas Größler,
Myrjam Stotz,
and
Nadine Schieritz
Mannheim University,
Germany


3. Agent-Based Simulation
Whereas in system dynamics the internal structure of a
system determines its dynamic tendencies, in the agent-
based simulation the dynamic behavior of a system arises
from the behavior of its elements, the agents, and the
interactions between them:

System Dynamics Agent-Based
Simulation
Basic building
block
Feedback loop Agent
Unit of analysis Structure Agents' rules
Level of modeling Macro Micro
Perspective Top-down Bottom-up

4. Scientific Problem
Agent-based simulation and system
dynamics use computer simulation to
investigate social and economic systems
characterized by non-linearity, delays and
feedback processes.
Both concentrate on understanding and
qualitative prediction of systems behavior.
 An integration of both approaches
might be fruitful.


5. The Software Used…
A software
framework for
creating agent
based simulations
using the Java
language
A visual modeling
tool for system
dynamic models
The Vensim DLL
allows you to call
Vensim functions
from other
applications


6. The Technical Integration
At least two Java classes have to be programmed:
the simulation environment and the agents’ class…
Java class of the
simulation
environment
• Based on a class
given by RePast
• Builds and
manages the agents
• Manages the
simulation
• Provides a
graphical user
interface
Java class of
the agents
• Represents the
mental model of
the agents
• Builds a Vensim
object to
communicate with
Vensim via the
Vensim DLL
Vensim object
(provided by
Vensim)
• Calls the Vensim
DLL
• Manages the
transfer of data and
commands
between the agent-
class and Vensim
provides creates creates calls


7. An example from Supply Chain
Management
Supplier
Agent
Manufacturer
Agent1
Manufacturer
Agent2
Manufacturer
Agent3
Behavior of the
manufacturer
RePast VensimDLL
Agent-based simulation System Dynamics
Behavior of the
manufacturer
Behavior of the
manufacturer


8. The Agents’ Behavior
Inventory
Acquisition Rate Shipment
Rate
Desired
Acquisition Rate
Acquisition
Adjustment from
Inventory
Desired
Inventory
Expected
Order Rate Change in
Exp Orders
Inventory
Adjustment
Time
Desired
Inventory
Coverage
Time to Average
Order Rate
Order
Fulfillment
Ratio
Table for Order
Fulfillment
Supply LineOrders Placed
Rate
Acquisition Lag
Adjustment for
Supply Line
Desired
Supply Line
Orders Placed
Supply Line
Adjustment Time
Customer
Order Rate
B
Order
Fulfillment
B Inventory Control
B
Supply Line
Control -
-
+
+
+
+
+
- -
+
+
-
+
+
+
-
Desired
Shipment
Rate
+
Maximum
Shipment
Rate
Minimum
Order
Processing
Time
+
+
-
-
+
Inventory
Coverage
+ -
Safety
Stock
Coverage
+
+
+
Customer
Orders
Backlog Backlog
Change Rate
+ -
+
Supplies
Received
Init Supply Line Init Inventory
Init Expected
Order Rate Init Customer
Orders Backlog
+
+
In each step of a simulation the agent-based modeled manufacturers
call this System Dynamics model (modified after Sterman 2000)...


9. The Simulation in RePast Start and stop
a simulation
Graphs represent
the results of
simulations
In a probe map
parameters of the
simulation can be set


10. Conclusions
 The presented software solution provides a
prototypical common technical platform to
examine problems that suggest the
integration of the two simulation concepts
 More flexibility is provided in modeling and
simulating dynamic systems (using in each
part of the model the method fitting best)
 The advantages of both methods can be used,
their disadvantages can be reduced
 Skills in Java are necessary


11. Further Research
 Extend and improve the model technically,
e.g. make simulations more user-friendly
 Use this platform to model other problems
that suggest the integration of the two
simulation concepts
 Use this software interface as a basis to model
an integration with System Dynamics on
macro-level and agent-based simulations on
micro-level
 Investigate effects of combined methods on
validity of models


12. References
 RePast: http://repast.sourceforge.net
 Venism: http://www.vensim.com
 Phelan, SE. 1999. A Note on the Correspondence between
Complexity and Systems Theory. Systemic Practice and
Action Research 12(3): 237–246.
 Schieritz, N, Größler, A. 2003. Emergent Structures in Supply
Chains: A Study Integrating Agent-Based and System
Dynamics Modeling. Proceedings of the 36th Hawaiian
International Conference on Systems Science, Wailea.
 Scholl, HJ. 2001a. Agent-based and System Dynamics
Modeling: A Call for Cross Study and Joint Research.
Proceedings of the 34th Hawaiian International Conference
on Systems Science, Wailea.
 Sterman, JD. 2000. Business Dynamics – Systems Thinking
and Modeling for a Complex World, Boston.
