Fishwick P.A. (editor) Handbook of Dynamic System Modeling

Подождите немного. Документ загружается.

8-8 Handbook of Dynamic System Modeling

Kernel

Subsimulators

GIS

Agent implementations

Viewer

Area

Buildings

Roads

Agent proxies

FIGURE 8.2 Architecture of a typical testbed for evaluating agents: The RoboCup Rescue simulator. (From Rescue

Technical Committee, T. R. (2000). RoboCup Rescue simulator manual. Ver. 0, Rev. 4.)

information access, and distributed control. Simulators are used to describe the dynamics of typical

disasters, e.g., such as the spread of ﬁre, the collapse of buildings, and blocking of highways. Rescue

complements RoboCup-Soccer with challenges in logistics, long-range planning, and collaboration among

heterogeneous agents. RCR takes a special position among test beds, as it is not only a challenging

artiﬁcial environment, but is directly aimed at socially signiﬁcant problems. Besides the ability to put

agent architectures to test with the Rescue simulator, RCR is aimed to make progress in emergency

decision support for large-scale disaster scenarios (Kitano et al., 1999). Furthermore, the RCR-simulator is

intended to function as a standard benchmark environment to compare and evaluate crisis management

and provide the ground for digitally empowered rescue brigades. Figure 8.2 depicts the structure of the

RCR-simulator (Rescue Technical Committee, 2000).

Each of these test beds supports testing of agents within predeﬁned, dynamic scenarios. They may

cast light on what is important and concentrate attention on key problems in a scientiﬁc area as they

operationalize scientiﬁc paradigms (Sim et al., 2003). However, as Hanks points out, testing should

ultimately be aimed at analyzing how agents will perform in the environment they are constructed for.

With agents, AI methods have moved toward embedded applications, “the ultimate interest being not

simpliﬁed systems and environments but rather real world systems deployed in complex environments”

(Hanks et al., 1993, p. 18). Consequently, recent benchmarks (Kitano and Tadokoro, 2001; Weyns et al.,

2005a) blur the border between artiﬁcial and real domains.

8.3.2 Simulation for Designing Multiagent Systems

Unlike testing in the small, where general strategies of agents are tested, this section discusses concrete

agents designed for concrete applications. Simulation is used for testing agents in a virtual environment.

Therefore, a valid model of the concrete environment the agent shall dwell in is required.

Generally, testing checks an implementation against some requirements. To this end, a set of test cases

has to be generated to cover the requirements as much as possible. If testing cannot be done exhaustively,

selecting suitable test cases becomes crucial. Efﬁcient testing relies on the use of explicit models representing

the conditions under which an implementation has to operate—its environment. Whereas most of the

Agent-Oriented Modeling in Simulation 8-9

testing approaches concentrate on models of the software itself, i.e., the speciﬁcation of requirements from

which test cases are derived, model-based testing uses additional explicit models for test case selection

(Pretschner et al., 2004). To constrain the set of test cases, model-based testing introduces a model of the

test scenario to distinguish signiﬁcant test cases. A random selection would reduce the number of test cases

arbitrarily. In contrast, the integration of further knowledge, e.g., about critical situations or common user

behavior, allows a goal-directed focusing of test cases. This is of particular importance for testing agent

software, as they are usually intended to operate in rather complex settings.

The usage of a virtual environment in contrast to the real environment typically reduces costs and

efforts and allows to test a system’s behavior in “rare event situations.” Virtual environments are easier

to observe and to control, and probe effects are easier to manage. Environment models can be used to

generate different test cases dynamically during simulation, including speciﬁc interaction patterns and

time constraints. Thereby, the validity of the environmental models is crucial, independent of whether

abstract models of agents are experimentally evaluated, single agent modules are embedded for testing, or

entire agent systems are plugged into the virtual environment (Röhl and Uhrmacher, 2005).

For example, have a look at peer-to-peer (P2P) networks (Moro et al., 2005), which differ from typical

client–server architectures in assigning to each node equivalent capabilities and responsibilities. P2P

networks are characterized by autonomous nodes and a dynamic topology as nodes might leave or enter

the network dynamically. For publishing and requesting information services in P2P networks, multiagent

software systems are designed (König-Ries et al., 2004). To evaluate different protocol strategies, classical

network simulators are considered as too inﬂexible and being on a too detailed level of abstraction (Klein,

2003). Therefore, simulators are developed from scratch, which simulate the underlying network at a

rather abstract level (Minar et al., 1999; Huebsch et al., 2003), e.g., by counting hops. Often, more detail is

required on the application level; e.g., instead of using simple “Churn”-rates, which describe the dynamics

of the topology (Joseph, 2003), the user’s individual activities are taken into account (König-Ries et al.,

2004). As those users can be modeled as agents, human behavior representation becomes part of the

environment software agents are tested in.

This has also been the case in developing a test environment for Autominder. Autominder is a distributed

monitoring and reminder system developed at the University of Michigan (Pollack et al., 2003). It is being

designed to support older adults with mild to moderate cognitive impairment in their activities of daily

living. The software is installed in the client’s house, together with a set of sensors and given a list of

actions that must be performed during the day. At runtime, Autominder evaluates sensor echoes from the

environment, reasons about ongoing user activity, and compares its ﬁndings to the given client plan to

detect forgotten actions and remind the user of their execution.

Autominder complements a human in-place nurse and thus has a huge responsibility for its client’s

safety. Consequently, it must be tested in a variety of application scenarios before its release. Since those

scenarios are not always safe for a human client and may be hard to observe during human-in-the-loop

ﬁeld tests, a virtual testing environment for Autominder has been developed (Gierke et al., 2006). Its

layout is shown in Figure 8.3. The testing environment focuses on the model of an elderly person who

acts as the Autominder user. Besides that, it includes a model of the client’s living environment where

Autominder is supposed to work. The Elderly model is designed as a human behavior representation that

represents an elderly person in terms of actions and mental state (Gierke and Uhrmacher, 2005). There-

fore, it provides two basic client functionalities: emulate suitable elderly behavior and react to incoming

reminders.

Depending on its initial memory ﬁtness and current stress level, the Elderly model may forget some of

the plan actions. If the model receives an external reminder, it interrupts this default behavior as reminders

should be evaluated with priority. However, there is no guarantee the Elderly model will cooperate with

Autominder and execute the requested action. Depending on its initial cooperativeness and current degree

of annoyance, the Elderly model may as well choose to ignore reminders.

In this scenario, we ﬁnd agents as a metaphor for modeling, i.e., part of the virtual environment, the

Elderly model, is described utilizing the agent metaphor HBR; and we ﬁnd agents as a subject for being

tested,i.e., the Autominder software. In manyenvironmentsfor which software agents are designed, human

8-10 Handbook of Dynamic System Modeling

Coop

Virtual environment

Rem REQ

Sens NTF

Elderly

Action

Planner

EIS

Exec NTF

Act NTF

Coorp REQ

Coorp RSP

Rem NTF

Frgt REQ

Frgt RSP

Autominder

I/O model

Simulator

Flat

FIGURE 8.3 Testing of an agent application (Autominder) by coupling it to a simulated environment. (From Gierke,

M., J. Himmelspach, M. Röhl, and A. Uhrmacher (2006). Modeling and simulation of tests for agents. In German

Conference on Multi-Agent System Technologies MATES’06, to appear.)

beings play an important role and thus have to be taken into account when designing virtual environments

for testing.

8.4 Conclusion

Agents are a metaphor used for designing software systems and for modeling dynamic systems that shall

be evaluated by simulation. Multiagent systems as communities of interacting autonomous entities are

supposed to function in dynamic environments. Thus, simulation is used as a means to evaluate and test

agents. Accordingly, the relationship that ties agents and simulation is multifaceted.

The objective of the chapter has been to entangle the relations between agents and modeling against the

background of simulation from two starting points: using the agent metaphor for modeling and modeling

the virtual environment for analyzing agents. Thereby, the focus moved from the structure of single agents

to their environments. These two approaches were further structured by the criteria whether analysis or

design was the objective that motivated the modeling effort. As it turned out, many dependencies between

these perspectives exist, which blur the distinction at some points. Figure 8.4 shows the introduced

dimensions in relating agents and simulation from a modeling perspective.

For example, if we look at robots playing soccer as a possible application, a soccer simulator would help

designing the software to run on the robots. However, current RoboCup simulators are not aimed toward

this goal. Therefore, their models would need to include more physical and sensory details of the robots

and the playing ground. Consequently, although simulators are used for designing soccer-playing robots,

these are quite different from the simulators used in the RoboCup-Soccer league illustrating quite nicely

that the objective inﬂuences the suitability of models. However, current approaches are aimed at enriching

the RoboCup-Soccer simulator for supporting the design of soccer-playing robots, thus, bridging the gap

between testing in the small and testing in the large, or simulation for understanding and simulation for

design.

Agent-Oriented Modeling in Simulation 8-11

Agent metaphor

for

understanding

Agent metaphor

for

designing

Simulation for

understanding

agent software

Simulation for

evaluating

agent software

Test bed is refined for

realistic environment

Agent software

becomes subject

of design

Environmental models

including HBR move

into focus of interest

Generalize scenario for reuse

Scenario turns

into test bed

Understanding interactions

and organizational structures

Importance of technical

aspects increases

Interest is turned

toward social

phenomena

FIGURE 8.4 Interdependencies between agents for simulation and simulation for agents.

To analyze the strategies of multiagent systems, scenarios from ecology and sociology have often

been used. Scenarios like RCR show that not only implemented strategies and the resulting emergent

phenomena shall be analyzed by a virtual environment, but it is also hoped to learn more about decision

structures in and the ability to cope with real disaster scenarios. In this case, modeling and simulation is

not only used for agents, i.e., as a test bed to get a better understanding about coordination strategies when

multiple heterogeneous agents are involved, but also the agent metaphor helps to understand dynamic

systems, e.g., when dealing with real disaster scenarios, where the agent metaphor serves for modeling and

simulating this system.

If technical systems like unguided vehicles in manufacturing systems are perceived as agents and

described in the simulation based on the agent metaphor, their implementation is likely to be agent-

based as well. Thus, as the implementation and design of the system progresses, we gradually turn from

the view of using the agent metaphor during simulation, toward using simulation for analyzing the imple-

mented agents. If, as in testing the agent software Autominder or information retrieval in P2P networks,

human factors play a role, agent as a metaphor for modeling humans naturally become part of the virtual

environment aimed at evaluating the agent software. In this context,the ideas of using agents as a metaphor

for modeling and using modeling (and simulation) to evaluate agents are combined.

To conclude, let us summarize our discussion:

1. Agents are used as a metaphor for modeling a dynamic system.

(a) The agent metaphor supports a better understanding of certain phenomena in social, ecological,

or biological communities. It typically enriches traditional individual and multilevel modeling

approaches by sophisticated internal decision models of the individuals, and thus helps to

analyze phenomena that are supposed to depend on decision processes and preference models

(e.g., urban disasters).

(b) Whereas in sociology, simulation serves a better understanding,manytechnical areas depend on

simulation for designing. In this area we ﬁnd the agent metaphor as well. If the technical system

can only be evaluated taking human behavior into account, agents are used as a metaphor to

mimic the behavior of the humans, e.g., in the HBR approach. The goal is to have a realistic

behavior pattern rather than to have a valid model of human decision and behavior processes,

e.g., in trafﬁc simulation. If the technical system shows characteristics of a multiagent system,

i.e., a community of autonomously and concurrently interacting entities, the agent metaphor

8-12 Handbook of Dynamic System Modeling

is used for describing this system, which brings it close to the relation described in Section 8.2.2

(e.g., manufacturing system).

2. The other dimension that relates agents and simulation is using simulation for evaluating software

agents. This turns the focus of modeling from the agent toward the environment.

(a) Small world testing presents a simplistic world in which properties of a concrete agent or a

simpliﬁed version of an agent can be analyzed in isolation, a better understanding of agents and

their behaviors being its purpose. Scenarios are often chosen by adopting social or biological

systems. However, less the understanding of the social society but more the implications of

the speciﬁc problem solving and cooperation strategies are the focus of interest (examples are

ILEWORLD, RoboCup simulation league).

(b) Testing in the large requires a valid model of the environment the agents are supposed to work

in. The focus is on evaluating design alternatives of concrete software agent applications. If the

environment comprises humans, they will likely be presented as agents. However, less with the

intention to understand humans in their interaction with software, but with the intention to

mimic the human behavior reasonably well to assess the performance of the software agent in

interaction with humans (examples are P2P systems and Autominder).

By focusing on the modeling level, research areas like using agents for realizing simulation systems, e.g.,

in supporting data-driven, online simulations (Low et al., 2005) have not been considered. Similarly, the

many challenges of an efﬁcient and effective execution of multiagent systems and their environments have

not been discussed, nor any of the steps toward addressing these challenges, e.g., by ﬂexible simulation

layers that support different types of executions and synchronization between agents and simulation

(Himmelspach and Uhrmacher, 2004), or e.g., by an efﬁcient handling of shared states in distributed,

parallel execution (Lees et al., 2005), nor were the different simulation tools that exist, most of which

originated in the realm of social simulations (Tobias and Hofmann, 2004), presented.

References

Agre, P. and D. Chapman (1987). PENGI: An implementation of a theory of activity. In Sixth National

Conference on Artiﬁcial Intelligence (AAAI-87), Seattle, WA, pp. 268–272.

Axtell, R. (2000). Why agents? On the varied motivations for agent computing in the social sciences.

Working Paper No. 17.

Balmer, M., K. Nagel, and B. Raney (2004). Large-scale multi-agent simulations for transportation

applications. ITS Journal 8, 1–17.

Batty, M. (2005). Cities and Complexity: Understanding Cities with Cellular Automata, Agent-Based Models,

and Fractals. London: MIT Press.

Baumgaertel, H., S. Brueckner, V. Parunak, R. Vanderbok, and J. Wilke (2003). Agent models of supply

network dynamics. In T. P. Harrison, H. L. Lee, and J. J. Neale (Eds.), The Practice of Supply Chain

Management, International Series on Operations Research and Management Science. New York:

Kluwer.

Brooks, R. A. (1991). Intelligence without reason. In J. Myopoulos and R. Reiter (Eds.), Proceedings

of the 12th International Joint Conference on Artiﬁcial Intelligence (IJCAI-91), Sydney, Australia,

pp. 569–595. Morgan Kaufmann: San Mateo, CA, USA.

Bunge, M. (1979). Ontology II: A World of Systems, Volume 4 of Treatise of Basic Philosophy. Dordrecht:

Reidel.

Casagni, M. and M. Lyell (2003). Comparison of two component frameworks: The ﬁpa-compliant multi-

agent system and the web-centre j2ee platform. In ICSE ’03: Proceedings of the 25th International

Conference on Software Engineering, Washington, DC, USA, pp. 341–351. IEEE Computer Society.

Chen, S., S. Ganguli, and C. A. Hunt (2004). An agent-based computational approach for represent-

ing aspects of in vitro multi-cellular tumor spheroid growth. In Proceedings of the 26th Annual

International Conference of the IEEE EMBS, pp. 691–694.

Agent-Oriented Modeling in Simulation 8-13

Clark, M., K. Irvig, and W. Wobcke (1997). Emergent properties of teams of agents in the Tileworld. In

L. Cavedon, A. Rao, and W. Wobcke (Eds.), Intelligent Agent Systems, Volume 1209 of Lecture Notes

in Artiﬁcial Intelligence, pp. 164–176. Berlin: Springer.

Conte, R. and C. Castelfranchi (Eds.) (1995). Cognitive and Social Action. London: UCL Press.

Conte, R., B. Edmonds, S. Moss, and R. K. Sawyer (2001). Sociology and social theory in agent based social

simulation: A symposium. Computational & Mathematical Organization Theory 7(3), 183–205.

David, N., J. S. Sichman, and H. Coelho (2003). Towards an emergence-driven software process for agent-

based simulation. In J. Sichman, F. Bousquet, and P. Davidsson (Eds.), MABS 2002, Volume 2581

of Lecture Notes in Artiﬁcial Intelligence, pp. 89–104. Berlin: Springer.

Doran, J. (1997). From computer simulation to artiﬁcial societies. SCS Transaction on Computer Simulation

14(2), 69–78.

Doran, J., N. Gilbert, U. Müller, and K. Troitzsch (Eds.) (1996). Object-Oriented and Agent-Oriented

Simulation – Implications for Social Science Applications, Lecture Notes in Economics and

Mathematical Systems, Berlin: Springer.

Doran, J., M. Palmer, N. Gilbert, and P. Mellars (1994). The EOS project: Modeling upper paleolithic social

change. In N. Gilbert and J. Doran (Eds.), Simulating Societies, pp. 195–223. London: UCL Press.

Edmonds, B. and M. Möhring (2005). Agent-based simulation modeling in social and organizational

domains. Simulation 81(3), 173–174.

Ephrati, E., M. Pollack, and S. Ur (1995). Deriving multi-agent coordination through ﬁltering strategies.

In 14th International Joint Conference on Artiﬁcial Intelligence, pp. 679–687. Morgan Kaufmann: San

Mateo, CA.

Etzioni, O. (1993). Intelligence without robots: A reply to brooks. AI Magazine 14(4), 7–13.

Ewert, U. C., M. Röhl, and A. M. Uhrmacher (2003). What good are deliberative interventions in

large scale disasters? Exploring the consequences of crisis management in pre-modern towns

with agent-oriented simulation. In Agent Based Computational Demography. Berlin: Physica Verlag

(Springer).

Ferber, J. (1999). Multi-Agent System: An Introduction to Distributed Artiﬁcial Intelligence. Boston, MA:

Addison-Wesley.

Gierke, M., J. Himmelspach, M. Röhl, and A. Uhrmacher (2006). Modeling and simulation of tests for

agents. In German Conference on Multi-Agent System Technologies MATES’06. LNCS Volume 4196,

pp. 49–60. Heidelberg: Springer.

Gierke, M. and A. M. Uhrmacher (2005). Modeling elderly behavior for simulation-based testing on agent

software. In Proceedings of the Conceptual Modeling and Simulation Conference (CMS 2005).San

Diego, CA: SCS.

Gilbert, N. (2005). When does social simulation need cognitive models? In R. Sun (Ed.), Cognition and

Multi-Agent Interaction: From Cognitive Modeling to Social Simulation. Cambridge: Cambridge

University Press.

Gilbert, N. and K. G. Troitzsch (2005). Simulation for the Social Scientist (2nd ed.). Maidenhead: Open

University Press.

Goldman, G. and J. Rosenschein (1994). Emergent coordination through the use of cooperative state-

changing rules. In National Conference on Artiﬁcial Intelligence, Seattle, WA, pp. 432–437. Menlo

Park, CA: American Association for Artiﬁcial Intelligence.

Hanks, S., M. E. Pollack, and P. R. Cohen (1993). Benchmarks, test beds, controlled experimentation and

the design of agent architectures. AAAI(Winter), 17–42.

Helbing, D. and K. Nagel (2004). The physics of trafﬁc and regional development. Contemporary Physics

45(5), 405–426.

Helleboogh, A., G. Vizzari, A. Uhrmacher, and F. Michel (2007). Modeling dynamic environments in

multi-agent simulation. Journal of Autonomous Agents and Multi-Agent Systems 14(1), 87–116.

Himmelspach, J. and A. M. Uhrmacher (2004). A component-based simulation layer for James. In PADS

’04: Proceedings of the Eighteenth Workshop on Parallel and Distributed Simulation, Kufstein, Austria,

pp. 115–122. New York: IEEE Computer Society Press.

8-14 Handbook of Dynamic System Modeling

Huebsch, R., J. M. Hellerstein, N. Lanham, B. T. Loo, S. Shenker, and I. Stoica (2003). Querying the

internet with pier. In VLDB, pp. 321–333.

Jennings, N. R., K. Sycara, and M. Wooldridge (1998). A roadmap of agent resarch and development.

Autonomous Agents and Multi-Agent Systems 1(1), 275–306.

Joseph, S. (2003). An extendible open source P2P simulator, P2P Journal, pp. 1–15.

Kinny, D. and M. Georgeff (1991). Commitment and effectiveness of situated agents. In Proceedings of the

International Joint Conference, pp. 82–88. IJCAI, Morgan Kaufmann: San Mateo, CA.

Kitano, H. and S. Tadokoro (2001). RoboCup rescue: A grand challenge for multiagent and intelligent

systems. AI Magazine 22(1), 39–52.

Kitano, H., S. Tadokoro, H. Noda, I. Matsubara, T. Takhasi, A. Shinjou, and S. Shimada (1999). Robocup-

rescue: Search and rescue for large scale disasters as a domain for multi-agent research. In Proceedings

of the IEEE Conference on Systems, Men, and Cybernetics.

Kitano, H., M. Tambe, P. Stone, M. Veloso, S. Coradeschi, E. Osawa, H. Matsubara, I. Noda, and M. Asada

(1997). The RoboCup synthetic agent challenge 1997. In International Joint Conference on Artiﬁcial

Intelligence IJCAI’97. Morgan Kaufmann: San Mateo, CA.

Klein, M. (2003). DIANEmu: A java based generic simulation environment for distributed protocols.

Technical Report 2003–7, University of Karlsruhe.

Klügl, F. and A. L. C. Bazzan (2004). Route decision behaviour in a commuting scenario: Simple heuristics

adaptation and effect of trafﬁc forecast. JASSS 7(1).

Knorr-Cetina, K. and A. Cicourel (Eds.) (1981). Advances in Social Theory and Methodology—Toward an

Integration of Micro and Macro Sociologies. Boston: Routledge and Kegan Paul.

König-Ries, B., M. Klein, and T. Breyer (2004). Activity-based user modeling in wireless net-

works. Mobile Networks and Applications. Special Issue on Internet Wireless Access: 802.11 and

Beyond.

Lees, M., B. Logan, R. Minson, T. Oguara, and G. Theodoropoulos (2005). Distributed simulation of MAS.

In P. Davidson, B. Logan, and K. Takadama (Eds.), AAMAS 2004, Volume 3415 of Lecture Notes in

Computer Science, pp. 25–36. Berlin: Springer.

Low, M. Y. H., K. W. Lye, P. Lendermann, S. J. Turner, R. T. W. Chim, and S. H. Leo (2005). An agent-based

approach for managing symbiotic simulation of semiconductor assembly and test operation. In

AAMAS ’05: Proceedings of the Fourth International Joint Conference on Autonomous Agents and

Multiagent Systems, New York, NY, pp. 85–92. ACM Press.

Luck, M., N. Grifﬁths, and M. d’Inverno (1997). From agent theory to agent construction: A case study.

In J. Mueller, M. Wooldridge, and N. Jennings (Eds.), Third International Workshop on Agent Theo-

ries, Architectures, and Languages, Volume 1193 of Lecture Notes in Artiﬁcial Intelligence, London,

pp. 49–63. Berlin: Springer.

Melo, F., R. Choren, R. Cerqueira, C. Lucena, and M. Blois (2004). Deploying agents with the CORBA

component model. In W. Emmerich and A. Wolf (Eds.), CD 2004, Volume 3083 of Lecture Notes

in Computer Science, pp. 234–247. Berlin: Springer.

Minar, N., K. H. Kramer, and P. Maes (1999). Cooperating mobile agents for dynamic network routing.

In A. L. G. Hayzelden and J. Bigham (Eds.), Software Agents for Future Communication Systems,

pp. 287–304. Heidelberg, Germany: Springer.

Moro, G., S. Bergamaschi, and K. Aberer (Eds.) (2005). Agents and Peer-to-Peer Computing, Volume 3601

of Lecture Notes in Computer Science. Berlin: Springer.

Negnevitsky, M. (2002). Artiﬁcial Intelligence: A Guide to Intelligent Systems. Harlow, England:

Addison-Wesley.

Odell, J. J. (2002). Objects and agents compared. Journal of Object Technology 1(1), 41–53.

O’Hare, G. M. P. and N. R. Jennings (Eds.) (1996). Foundations of Distributed Artiﬁcial Intelligence.

New York, NY: Wiley.

Parunak, H. V. D. (1997). “Go to the ant”: Engineering principles from natural agent systems. Annals of

Operation Research 75, 69–101.

Agent-Oriented Modeling in Simulation 8-15

Parunak, H. V. D. (1999). From chaos to commerce: Practical issues and research opportunities in the

nonlinear dynamics of decentralized manufacturing systems. In Proceedings of Second International

Workshop on Intelligent Manufacturing Systems, pp. k15–k25.

Pew, R. W. and A. S. Mavor (Eds.) (1997). Representing Human Behavior in Military Simulations: Interim

Report. Washington, DC: National Academy Press.

Pew, R. W. and A. S. Mavor (Eds.) (1998). Modeling Human and Organizational Behavior: Application to

Military Simulations. Washington, DC: National Academy Press.

Philips, A. and J. Bresina (1991). NASA TILEWORLD manual. Technical Report TR_FIA_91-04, MASA

Ames Research Center, Mountain View, CA.

Pollack, M. E., L. Brown, D. Colbry, C. E. McCarthy, C. Orosz, B. Peintner, S. Ramakrishnan, and

I. Tsamardinos (2003). Autominder: An intelligent cognitive orthotic system for people with

memory impairment. Robotics and Autonomous Systems 44, 273–282.

Pollack, M. E. and M. Ringuette (1990). Introducing the TILEWORLD: Experimentally evaluating agent

architectures. In AAAI-90, Boston, MA, pp. 183–189.

Pretschner, A., O. Slotosch, E. Aiglstorfer, and S. Kriebel (2004). Model-based testing for real—The

inhouse card case study. International Journal on Software Tools for Technology Transfer (STTT)

5(2–3), 140–157.

Rao, A. S. and M. P. Georgeff (1995). BDI agents: From theory to practice. Technical Report 56, Australian

Artiﬁcial Intelligence Institute.

Rescue Technical Committee, T. R. (2000). RoboCup Rescue simulator manual. Ver. 0, Rev. 4.

Röhl, M. and A. M. Uhrmacher (2005). Controlled experimentation with agents—Models and implemen-

tations. In M.-P. Gleizes, A. Omicini, and F. Zambonelli (Eds.), Post-Proceedings of the 5th Workshop

on Engineering Societies in the Agents World, Volume 3451 of Lecture Notes in Artiﬁcial Intelligence,

pp. 292–304. Berlin: Springer.

Russell, S. J. and P. Norvig (2003). Artiﬁcial Intelligence: A Modern Approach (2nd ed.). New Jersey:

Prentice-Hall.

Shoham, Y. (1993). Agent-oriented programming. Artiﬁcial Intelligence 60, 51–92.

Silverman, B. G. (2003). Metrics and Methods in Human Performance Research toward Individual and Small

Unit Simulation, Chapter 9 Human Performance Simulation. Washington, DC, Human Systems

Information Analysis Center.

Sim, S. E., S. Easterbrook, and R. C. Holt (2003). Using benchmarking to advance research: A challenge

to software engineering. In ICSE ’03: Proceedings of the 25th International Conference on Software

Engineering, Washington, DC, USA, pp. 74–83. IEEE Computer Society.

Tambe, M., J. Adibi,Y. Alonaizon, A. Erdem, G. Kaminka, S. Marsella, and I. Muslea (1999). Building agent

teams using an explicit teamwork model and learning. Artiﬁcial Intelligence 110(2), 215–239.

Tanenbaum, A. S. and M. van Steen (2002). Distributed Systems: Principles and Paradigms.NewYork:

Prentice-Hall.

Tobias, R. and C. Hofmann (2004). Evaluation of free java-libraries for social-scientiﬁc agent based

simulation. Journal of Artiﬁcial Societies and Social Simulation 7(1).

Uhrmacher, A., P. Fishwick, and B. Zeigler (Eds.) (2001). Special Issue: Agents in Modeling and Simulation;

Exploiting the Metaphor, Volume 89 of Proceedings of the IEEE, pp. 127–213.

Uhrmacher, A. M. (1997). Concepts of object- and agent-oriented simulation. Transactions on SCS 14(2),

59–67.

Weiss, G. (Ed.) (2000). Multiagent Systems: A Modern Approach to Distributed Artiﬁcial Intelligence.

Cambridge, MA: MIT Press.

Weyns, D., A. Helleboogh, and T. Holvoet (2005a). The packet-world: A test bed for investigating situ-

ated multi-agent systems. In R. Unland, M. Klusch, and M. Calisti (Eds.), Software Agent-Based

Applications, Platforms and Development Kits, Whitestein Series in Software Agent Technology.

Weyns, D., K. Schelfthout, T. Holvoet, and T. Lefever (2005b). Decentralized control of E’GV

transportation systems. In Proceedings of AAMAS’05, pp. 67–74. ACM Press.

8-16 Handbook of Dynamic System Modeling

Wise, B. P., M. McDonald, L. M. Reuss, and J. Aronson (2001). Task order (TO) 69: ATM human

behavior modeling approach study. Technical report, National Air and Space Administration

(NASA), Arlington, VA. Technical Advance in Air Transportation Concepts and Technologies

(AATT).

Wooldridge, M. (2000). Multiagent Systems: A Modern Approach to Distributed Artiﬁcial Intelligence,

Chapter Intelligent Agents. Cambridge, MA: MIT Press.

Wooldridge, M. (2002). An Introduction to MultiAgent Systems. Chichester, England: Wiley.

Wooldridge, M. and N. Jennings (1998). Pitfalls of agent-oriented development. In Proceedings of the 2nd

International Conference on Autonomous Agents (Agents-98), Minneapolis.

Wooldridge, M. and D. Vandekerckhove (1994). MyWorld: An agent-oriented testbed for distributed

artiﬁcial intelligence. In S. Deen (Ed.), CKBS-93 Proceedings of the 1993 Special Interest Group on

Cooperating Knowledge-Based Systems, Dake Centre, University of Keele, UK.

Yilmaz, L. and T. Ören (2005). Special issue on agent-directed simulation. Simulation 81(7), 463–464.

Distributed Modeling

SimonJ.E.Taylor

Brunel University

9.1 Introduction .....................................................................9-1

9.2 Modeling with COTS Simulation Packages ....................9-2

9.3 Distributed Simulation.....................................................9-3

9.4 CSP-Based Distributed Simulation .................................9-5

The Problem of CSP-Based Distributed Simulation

•

CSP-Based Distributed Simulation: Current Progress

9.5 A Standards-Based Approach ..........................................9-7

The Simulation Interoperability Standards Organization

•

Emerging Standards and the CSPI-PDG

•

TheTypeI

Interoperability Reference Model

•

The Entity

Transfer Speciﬁcation

9.6 Case Study ........................................................................9-13

Illustrative Protocol

9.7 Conclusion .......................................................................9-16

9.1 Introduction

Other chapters in this book discuss various aspects of dynamic systems modeling. The purpose of this

chapter is to present an introduction to some of the contemporary innovations in the use of distributed

computing techniques to support modeling. This is called distributed simulation and can be deﬁned as

the distribution of the execution of a single run of a simulation program across multiple processors

(Fujimoto, 2000). The various motivations for this include: the reduction of the execution time of a

single simulation run, the use of multiple computers to support the memory needs of the simulation

when one computer cannot, and the linking of simulations sited in different locations (Fujimoto, 2003).

A cursory examination of the many Winter Simulation Conferences (WSC) (www.informs-cs.org and

www.acm.org/dl), the Principles of Advanced and Distributed Simulation (PADS) (www.acm.org/dl), and

Simulation Interoperability Workshops (SIW) (www.sisostds.org) show the wide applications and issues

of distributed simulation. By way of introduction, and to give focus to this work, we restrict our discussion

to the use of distributed simulation in the ﬁeld of modeling associated with the use of Commercial-off-the-

shelf (COTS) Simulation Packages (CSPs). CSPs are widely used in industry. In this chapter, we term the

combination of distributed simulation and CSPs, CSP-based distributed simulation.

In addition to the above general reasons to use distributed simulation, additional reasons to use

CSP-based distributed simulation include the interoperation of discrete-event simulations across virtual

organizations, extended enterprises, and supply chains; the reduction of the cost of model development

by enabling the reuse of distributed model components; and the protection of intellectual property (infor-

mation hiding in distributed models) (Gan et al., 2000; Mertins and Rabe, 2002; Paul and Taylor, 2002;

Robinson et al., 2004). Additionally, variants of distributed simulation techniques can also reduce the

time taken for simulation experimentation (distributed replication and experimentation) and reduce sim-

ulation project costs (remote model execution, group working) (Robinson, 2005; Taylor et al., 2005).

9-1